Computer Vision and Pattern Recognition (cs.CV)

Abstract: Automatic surgical phase recognition is one of the key technologies to support Video-Based Assessment (VBA) systems for surgical education. Utilizing temporal information is crucial for surgical phase recognition, hence various recent approaches extract frame-level features to conduct full video temporal modeling. For better temporal modeling, we propose SlowFast Temporal Modeling Network (SF-TMN) for surgical phase recognition that can not only achieve frame-level full video temporal modeling but also achieve segment-level full video temporal modeling. We employ a feature extraction network, pre-trained on the target dataset, to extract features from video frames as the training data for SF-TMN. The Slow Path in SF-TMN utilizes all frame features for frame temporal modeling. The Fast Path in SF-TMN utilizes segment-level features summarized from frame features for segment temporal modeling. The proposed paradigm is flexible regarding the choice of temporal modeling networks. We explore MS-TCN and ASFormer models as temporal modeling networks and experiment with multiple combination strategies for Slow and Fast Paths. We evaluate SF-TMN on Cholec80 surgical phase recognition task and demonstrate that SF-TMN can achieve state-of-the-art results on all considered metrics. SF-TMN with ASFormer backbone outperforms the state-of-the-art Not End-to-End(TCN) method by 2.6% in accuracy and 7.4% in the Jaccard score. We also evaluate SF-TMN on action segmentation datasets including 50salads, GTEA, and Breakfast, and achieve state-of-the-art results. The improvement in the results shows that combining temporal information from both frame level and segment level by refining outputs with temporal refinement stages is beneficial for the temporal modeling of surgical phases.

Abstract: In this work, we proposed a new style-diverse dataset for the domain of motion style transfer. The motion dataset uses an industrial-standard human bone structure and thus is industry-ready to be plugged into 3D characters for many projects. We claim the challenges in motion style transfer and encourage future work in this domain by releasing the proposed motion dataset to the public. We conduct a comprehensive study on motion style transfer in the experiment using the state-of-the-art method, and the results show the proposed dataset's validity for the motion style transfer task.

Abstract: Autonomous driving presents many challenges due to the large number of scenarios the autonomous vehicle (AV) may encounter. End-to-end deep learning models are comparatively simplistic models that can handle a broad set of scenarios. However, end-to-end models require large amounts of diverse data to perform well. This paper presents a novel deep neural network that performs image-to-steering path navigation that helps with the data problem by adding one-shot learning to the system. Presented with a previously unseen path, the vehicle can drive the path autonomously after being shown the path once and without model retraining. In fact, the full path is not needed and images of the road junctions is sufficient. In-vehicle testing and offline testing are used to verify the performance of the proposed navigation and to compare different candidate architectures.

Abstract: Occlusion problem remains a key challenge in Optical Flow Estimation (OFE) despite the recent significant progress brought by deep learning in the field. Most existing deep learning OFE methods, especially those based on two frames, cannot properly handle occlusions, in part because there is no significant feature similarity in occluded regions. The multi-frame settings have the potential to mitigate the occlusion issue in OFE. However, the problem of Multi-frame OFE (MOFE) remains underexplored, and the limited works are specially designed for pyramid backbones and obtain the aligned temporal information by time-consuming backward flow calculation or non-differentiable forward warping transformation. To address these shortcomings, we propose an efficient MOFE framework named SplatFlow, which is realized by introducing the differentiable splatting transformation to align the temporal information, designing a One-to-Many embedding method to densely guide the current frame's estimation, and further remodelling the existing two-frame backbones. The proposed SplatFlow is very efficient yet more accurate as it is able to handle occlusions properly. Extensive experimental evaluations show that our SplatFlow substantially outperforms all published methods on KITTI2015 and Sintel benchmarks. Especially on Sintel benchmark, SplatFlow achieves errors of 1.12 (clean pass) and 2.07 (final pass), with surprisingly significant 19.4% and 16.2% error reductions from the previous best results submitted, respectively. Code is available at https://github.com/wwsource/SplatFlow.

Abstract: While VideoQA Transformer models demonstrate competitive performance on standard benchmarks, the reasons behind their success remain unclear. Do these models jointly capture and leverage the rich multimodal structures and dynamics from video and text? Or are they merely exploiting shortcuts to achieve high scores? We analyze this with $\textit{QUAG}$ (QUadrant AveraGe), a lightweight and non-parametric probe that systematically ablates the model's coupled multimodal understanding during inference. Surprisingly, QUAG reveals that the models manage to maintain high performance even when injected with multimodal sub-optimality. Additionally, even after replacing self-attention in multimodal fusion blocks with "QUAG-attention", a simplistic and less-expressive variant of self-attention, the models maintain high performance. This means that current VideoQA benchmarks and their metrics do not penalize shortcuts that discount joint multimodal understanding. Motivated by this, we propose the $\textit{CLAVI}$ (Counterfactual in LAnguage and VIdeo) benchmark, a diagnostic dataset for benchmarking coupled multimodal understanding in VideoQA through counterfactuals. CLAVI consists of temporal questions and videos that are augmented to curate balanced counterfactuals in language and video domains. Hence, it incentivizes, and identifies the reliability of learnt multimodal representations. We evaluate CLAVI and find that models achieve high performance on multimodal shortcut instances, but have very poor performance on the counterfactuals. Hence, we position CLAVI as a litmus test to identify, diagnose and improve the sub-optimality of learnt multimodal VideoQA representations which the current benchmarks are unable to assess.

Abstract: Pre-trained multi-modal vision-language models (VLMs) are becoming increasingly popular due to their exceptional performance on downstream vision applications, particularly in the few- and zero-shot settings. However, selecting the best-performing VLM for some downstream applications is non-trivial, as it is dataset and task-dependent. Meanwhile, the exhaustive evaluation of all available VLMs on a novel application is not only time and computationally demanding but also necessitates the collection of a labeled dataset for evaluation. As the number of open-source VLM variants increases, there is a need for an efficient model selection strategy that does not require access to a curated evaluation dataset. This paper proposes a novel task and benchmark for efficiently evaluating VLMs' zero-shot performance on downstream applications without access to the downstream task dataset. Specifically, we introduce a new task LOVM: Language-Only Vision Model Selection, where methods are expected to perform both model selection and performance prediction based solely on a text description of the desired downstream application. We then introduced an extensive LOVM benchmark consisting of ground-truth evaluations of 35 pre-trained VLMs and 23 datasets, where methods are expected to rank the pre-trained VLMs and predict their zero-shot performance.

Abstract: Faithfully reconstructing 3D geometry and generating novel views of scenes are critical tasks in 3D computer vision. Despite the widespread use of image augmentations across computer vision applications, their potential remains underexplored when learning neural rendering methods (NRMs) for 3D scenes. This paper presents a comprehensive analysis of the use of image augmentations in NRMs, where we explore different augmentation strategies. We found that introducing image augmentations during training presents challenges such as geometric and photometric inconsistencies for learning NRMs from images. Specifically, geometric inconsistencies arise from alterations in shapes, positions, and orientations from the augmentations, disrupting spatial cues necessary for accurate 3D reconstruction. On the other hand, photometric inconsistencies arise from changes in pixel intensities introduced by the augmentations, affecting the ability to capture the underlying 3D structures of the scene. We alleviate these issues by focusing on color manipulations and introducing learnable appearance embeddings that allow NRMs to explain away photometric variations. Our experiments demonstrate the benefits of incorporating augmentations when learning NRMs, including improved photometric quality and surface reconstruction, as well as enhanced robustness against data quality issues, such as reduced training data and image degradations.

Abstract: Masked autoencoder (MAE) has emerged as a promising self-supervised pretraining technique to enhance the representation learning of a neural network without human intervention. To adapt MAE onto volumetric medical images, existing methods exhibit two challenges: first, the global information crucial for understanding the clinical context of the holistic data is lacked; second, there was no guarantee of stabilizing the representations learned from the randomly masked inputs. To tackle these limitations, we proposed Global-Local Masked AutoEncoder (GL-MAE), a simple yet effective self-supervised pre-training strategy. GL-MAE reconstructs both the masked global and masked local volumes, which enables learning the essential local details as well as the global context. We further introduced global-to-global consistency and local-to-global correspondence via global-guided consistency learning to enhance and stabilize the representation learning of the masked volumes. Finetuning results on multiple datasets illustrate the superiority of our method over other state-of-the-art self-supervised algorithms, demonstrating its effectiveness on versatile volumetric medical image segmentation tasks, even when annotations are scarce. Codes and models will be released upon acceptance.

Abstract: Change detection using earth observation data plays a vital role in quantifying the impact of disasters in affected areas. While data sources like Sentinel-2 provide rich optical information, they are often hindered by cloud cover, limiting their usage in disaster scenarios. However, leveraging pre-disaster optical data can offer valuable contextual information about the area such as landcover type, vegetation cover, soil types, enabling a better understanding of the disaster's impact. In this study, we develop a model to assess the contribution of pre-disaster Sentinel-2 data in change detection tasks, focusing on disaster-affected areas. The proposed Context-Aware Change Detection Network (CACDN) utilizes a combination of pre-disaster Sentinel-2 data, pre and post-disaster Sentinel-1 data and ancillary Digital Elevation Models (DEM) data. The model is validated on flood and landslide detection and evaluated using three metrics: Area Under the Precision-Recall Curve (AUPRC), Intersection over Union (IoU), and mean IoU. The preliminary results show significant improvement (4\%, AUPRC, 3-7\% IoU, 3-6\% mean IoU) in model's change detection capabilities when incorporated with pre-disaster optical data reflecting the effectiveness of using contextual information for accurate flood and landslide detection.

Abstract: UAV distance estimation plays an important role for path planning of swarm UAVs and collision avoidance. However, the lack of annotated data seriously hinder the related studies. In this paper, we build and present a UAVDE dataset for UAV distance estimation, in which distance between two UAVs is obtained by UWB sensors. During experiments, we surprisingly observe that the commonly used stereo triangulation can not stand for UAV scenes. The core reason is the position deviation issue of UAVs due to long shooting distance and camera vibration, which is common in UAV scenes. To tackle this issue, we propose a novel position correction module (PCM), which can directly predict the offset between the image positions and the actual ones of UAVs and perform calculation compensation in stereo triangulation. Besides, to further boost performance on hard samples, we propose a dynamic iterative correction mechanism, which is composed of multiple stacked PCMs and a gating mechanism to adaptively determine whether further correction is required according to the difficulty of data samples. Consequently, the position deviation issue can be effectively alleviated. We conduct extensive experiments on UAVDE, and our proposed method can achieve a 38.84% performance improvement, which demonstrates its effectiveness and superiority. The code and dataset would be released.

Abstract: The Segmentation Anything Model (SAM) has recently emerged as a foundation model for addressing image segmentation. Owing to the intrinsic complexity of medical images and the high annotation cost, the medical image segmentation (MIS) community has been encouraged to investigate SAM's zero-shot capabilities to facilitate automatic annotation. Inspired by the extraordinary accomplishments of interactive medical image segmentation (IMIS) paradigm, this paper focuses on assessing the potential of SAM's zero-shot capabilities within the IMIS paradigm to amplify its benefits in the MIS domain. Regrettably, we observe that SAM's vulnerability to prompt forms (e.g., points, bounding boxes) becomes notably pronounced in IMIS. This leads us to develop a framework that adaptively offers suitable prompt forms for human experts. We refer to the framework above as temporally-extended prompts optimization (TEPO) and model it as a Markov decision process, solvable through reinforcement learning. Numerical experiments on the standardized benchmark BraTS2020 demonstrate that the learned TEPO agent can further enhance SAM's zero-shot capability in the MIS context.

Abstract: Quantization and pruning are known to be two effective Deep Neural Networks model compression methods. In this paper, we propose Automatic Prune Binarization (APB), a novel compression technique combining quantization with pruning. APB enhances the representational capability of binary networks using a few full-precision weights. Our technique jointly maximizes the accuracy of the network while minimizing its memory impact by deciding whether each weight should be binarized or kept in full precision. We show how to efficiently perform a forward pass through layers compressed using APB by decomposing it into a binary and a sparse-dense matrix multiplication. Moreover, we design two novel efficient algorithms for extremely quantized matrix multiplication on CPU, leveraging highly efficient bitwise operations. The proposed algorithms are 6.9x and 1.5x faster than available state-of-the-art solutions. We perform an extensive evaluation of APB on two widely adopted model compression datasets, namely CIFAR10 and ImageNet. APB shows to deliver better accuracy/memory trade-off compared to state-of-the-art methods based on i) quantization, ii) pruning, and iii) combination of pruning and quantization. APB outperforms quantization also in the accuracy/efficiency trade-off, being up to 2x faster than the 2-bits quantized model with no loss in accuracy.

Abstract: In this technical report, we present the solution developed by our team VIELab-HUST for text recognition through atmospheric turbulence in Track 2.1 of the CVPR 2023 UG$^{2}$+ challenge. Our solution involves an efficient multi-stage framework that restores a high-quality image from distorted frames. Specifically, a frame selection algorithm based on sharpness is first utilized to select the sharpest set of distorted frames. Next, each frame in the selected frames is aligned to suppress geometric distortion through optical-flow-based image registration. Then, a region-based image fusion method with DT-CWT is utilized to mitigate the blur caused by the turbulence. Finally, a learning-based deartifacts method is applied to remove the artifacts in the fused image, generating a high-quality outuput. Our framework can handle both hot-air text dataset and turbulence text dataset provided in the final testing phase and achieved 1st place in text recognition accuracy. Our code will be available at https://github.com/xsqhust/Turbulence_Removal.

Abstract: Learning effective spectral-spatial features is important for the hyperspectral image (HSI) classification task, but the majority of existing HSI classification methods still suffer from modeling complex spectral-spatial relations and characterizing low-level details and high-level semantics comprehensively. As a new class of record-breaking generative models, diffusion models are capable of modeling complex relations for understanding inputs well as learning both high-level and low-level visual features. Meanwhile, diffusion models can capture more abundant features by taking advantage of the extra and unique dimension of timestep t. In view of these, we propose an unsupervised spectral-spatial feature learning framework based on the diffusion model for HSI classification for the first time, named Diff-HSI. Specifically, we first pretrain the diffusion model with unlabeled HSI patches for unsupervised feature learning, and then exploit intermediate hierarchical features from different timesteps for classification. For better using the abundant timestep-wise features, we design a timestep-wise feature bank and a dynamic feature fusion module to construct timestep-wise features, adaptively learning informative multi-timestep representations. Finally, an ensemble of linear classifiers is applied to perform HSI classification. Extensive experiments are conducted on three public HSI datasets, and our results demonstrate that Diff-HSI outperforms state-of-the-art supervised and unsupervised methods for HSI classification.

Abstract: Robustly and accurately localizing objects in real-world environments can be challenging due to noisy data, hardware limitations, and the inherent randomness of physical systems. To account for these factors, existing works estimate the aleatoric uncertainty of object detectors by modeling their localization output as a Gaussian distribution $\mathcal{N}(\mu,\,\sigma^{2})\,$, and training with loss attenuation. We identify three aspects that are unaddressed in the state of the art, but warrant further exploration: (1) the efficient and mathematically sound propagation of $\mathcal{N}(\mu,\,\sigma^{2})\,$ through non-linear post-processing, (2) the calibration of the predicted uncertainty, and (3) its interpretation. We overcome these limitations by: (1) implementing loss attenuation in EfficientDet, and proposing two deterministic methods for the exact and fast propagation of the output distribution, (2) demonstrating on the KITTI and BDD100K datasets that the predicted uncertainty is miscalibrated, and adapting two calibration methods to the localization task, and (3) investigating the correlation between aleatoric uncertainty and task-relevant error sources. Our contributions are: (1) up to five times faster propagation while increasing localization performance by up to 1\%, (2) up to fifteen times smaller expected calibration error, and (3) the predicted uncertainty is found to correlate with occlusion, object distance, detection accuracy, and image quality.

Abstract: To be widely adopted, 3D facial avatars need to be animated easily, realistically, and directly, from speech signals. While the best recent methods generate 3D animations that are synchronized with the input audio, they largely ignore the impact of emotions on facial expressions. Instead, their focus is on modeling the correlations between speech and facial motion, resulting in animations that are unemotional or do not match the input emotion. We observe that there are two contributing factors resulting in facial animation - the speech and the emotion. We exploit these insights in EMOTE (Expressive Model Optimized for Talking with Emotion), which generates 3D talking head avatars that maintain lip sync while enabling explicit control over the expression of emotion. Due to the absence of high-quality aligned emotional 3D face datasets with speech, EMOTE is trained from an emotional video dataset (i.e., MEAD). To achieve this, we match speech-content between generated sequences and target videos differently from emotion content. Specifically, we train EMOTE with additional supervision in the form of a lip-reading objective to preserve the speech-dependent content (spatially local and high temporal frequency), while utilizing emotion supervision on a sequence-level (spatially global and low frequency). Furthermore, we employ a content-emotion exchange mechanism in order to supervise different emotion on the same audio, while maintaining the lip motion synchronized with the speech. To employ deep perceptual losses without getting undesirable artifacts, we devise a motion prior in form of a temporal VAE. Extensive qualitative, quantitative, and perceptual evaluations demonstrate that EMOTE produces state-of-the-art speech-driven facial animations, with lip sync on par with the best methods while offering additional, high-quality emotional control.

Abstract: Semantic scene completion (SSC) is crucial for holistic 3D scene understanding by jointly estimating semantics and geometry from sparse observations. However, progress in SSC, particularly in autonomous driving scenarios, is hindered by the scarcity of high-quality datasets. To overcome this challenge, we introduce SSCBench, a comprehensive benchmark that integrates scenes from widely-used automotive datasets (e.g., KITTI-360, nuScenes, and Waymo). SSCBench follows an established setup and format in the community, facilitating the easy exploration of the camera- and LiDAR-based SSC across various real-world scenarios. We present quantitative and qualitative evaluations of state-of-the-art algorithms on SSCBench and commit to continuously incorporating novel automotive datasets and SSC algorithms to drive further advancements in this field. Our resources are released on https://github.com/ai4ce/SSCBench.

Abstract: Image restoration under adverse weather conditions (e.g., rain, snow and haze) is a fundamental computer vision problem and has important indications for various downstream applications. Different from early methods that are specially designed for specific type of weather, most recent works tend to remove various adverse weather effects simultaneously through either spatial feature representation learning or semantic information embedding. Inspired by the various successful applications of large-scale pre-trained models (e.g, CLIP), in this paper, we explore the potential benefits of them for this task through both spatial feature representation learning and semantic information embedding aspects: 1) for spatial feature representation learning, we design a Spatially-Adaptive Residual (\textbf{SAR}) Encoder to extract degraded areas adaptively. To facilitate its training, we propose a Soft Residual Distillation (\textbf{CLIP-SRD}) strategy to transfer the spatial knowledge from CLIP between clean and adverse weather images; 2) for semantic information embedding, we propose a CLIP Weather Prior (\textbf{CWP}) embedding module to make the network handle different weather conditions adaptively. This module integrates the sample specific weather prior extracted by CLIP image encoder together with the distribution specific information learned by a set of parameters, and embeds them through a cross attention mechanism. Extensive experiments demonstrate that our proposed method can achieve state-of-the-art performance under different and challenging adverse weather conditions. Code will be made available.

Abstract: We propose a method for annotating videos of complex multi-object scenes with a globally-consistent 3D representation of the objects. We annotate each object with a CAD model from a database, and place it in the 3D coordinate frame of the scene with a 9-DoF pose transformation. Our method is semi-automatic and works on commonly-available RGB videos, without requiring a depth sensor. Many steps are performed automatically, and the tasks performed by humans are simple, well-specified, and require only limited reasoning in 3D. This makes them feasible for crowd-sourcing and has allowed us to construct a large-scale dataset by annotating real-estate videos from YouTube. Our dataset CAD-Estate offers 108K instances of 12K unique CAD models placed in the 3D representations of 21K videos. In comparison to Scan2CAD, the largest existing dataset with CAD model annotations on real scenes, CAD-Estate has 8x more instances and 4x more unique CAD models. We showcase the benefits of pre-training a Mask2CAD model on CAD-Estate for the task of automatic 3D object reconstruction and pose estimation, demonstrating that it leads to improvements on the popular Scan2CAD benchmark. We will release the data by mid July 2023.

Abstract: Large-scale visual localization systems continue to rely on 3D point clouds built from image collections using structure-from-motion. While the 3D points in these models are represented using local image features, directly matching a query image's local features against the point cloud is challenging due to the scale of the nearest-neighbor search problem. Many recent approaches to visual localization have thus proposed a hybrid method, where first a global (per image) embedding is used to retrieve a small subset of database images, and local features of the query are matched only against those. It seems to have become common belief that global embeddings are critical for said image-retrieval in visual localization, despite the significant downside of having to compute two feature types for each query image. In this paper, we take a step back from this assumption and propose Constrained Approximate Nearest Neighbors (CANN), a joint solution of k-nearest-neighbors across both the geometry and appearance space using only local features. We first derive the theoretical foundation for k-nearest-neighbor retrieval across multiple metrics and then showcase how CANN improves visual localization. Our experiments on public localization benchmarks demonstrate that our method significantly outperforms both state-of-the-art global feature-based retrieval and approaches using local feature aggregation schemes. Moreover, it is an order of magnitude faster in both index and query time than feature aggregation schemes for these datasets. Code will be released.

Abstract: Explainable AI aims to render model behavior understandable by humans, which can be seen as an intermediate step in extracting causal relations from correlative patterns. Due to the high risk of possible fatal decisions in image-based clinical diagnostics, it is necessary to integrate explainable AI into these safety-critical systems. Current explanatory methods typically assign attribution scores to pixel regions in the input image, indicating their importance for a model's decision. However, they fall short when explaining why a visual feature is used. We propose a framework that utilizes interpretable disentangled representations for downstream-task prediction. Through visualizing the disentangled representations, we enable experts to investigate possible causation effects by leveraging their domain knowledge. Additionally, we deploy a multi-path attribution mapping for enriching and validating explanations. We demonstrate the effectiveness of our approach on a synthetic benchmark suite and two medical datasets. We show that the framework not only acts as a catalyst for causal relation extraction but also enhances model robustness by enabling shortcut detection without the need for testing under distribution shifts.

Abstract: The process of transforming a raster image into a vector representation is known as image tracing. This study looks into several processing methods that include high-pass filtering, autoencoding, and vectorization to extract an abstract representation of an image. According to the findings, rebuilding an image with autoencoders, high-pass filtering it, and then vectorizing it can represent the image more abstractly while increasing the effectiveness of the vectorization process.

Abstract: This technical report introduces the winning solution of the team \textit{Segment Any Anomaly} for the CVPR2023 Visual Anomaly and Novelty Detection (VAND) challenge. Going beyond uni-modal prompt, \textit{e.g.}, language prompt, we present a novel framework, \textit{i.e.}, Segment Any Anomaly + (SAA$+$), for zero-shot anomaly segmentation with multi-modal prompts for the regularization of cascaded modern foundation models. Inspired by the great zero-shot generalization ability of foundation models like Segment Anything, we first explore their assembly (SAA) to leverage diverse multi-modal prior knowledge for anomaly localization. Subsequently, we further introduce multimodal prompts (SAA$+$) derived from domain expert knowledge and target image context to enable the non-parameter adaptation of foundation models to anomaly segmentation. The proposed SAA$+$ model achieves state-of-the-art performance on several anomaly segmentation benchmarks, including VisA and MVTec-AD, in the zero-shot setting. We will release the code of our winning solution for the CVPR2023 VAND challenge at \href{Segment-Any-Anomaly}{https://github.com/caoyunkang/Segment-Any-Anomaly} \footnote{The extended-version paper with more details is available at ~\cite{cao2023segment}.}

Abstract: Indoor rooms are among the most common use cases in 3D scene understanding. Current state-of-the-art methods for this task are driven by large annotated datasets. Room layouts are especially important, consisting of structural elements in 3D, such as wall, floor, and ceiling. However, they are difficult to annotate, especially on pure RGB video. We propose a novel method to produce generic 3D room layouts just from 2D segmentation masks, which are easy to annotate for humans. Based on these 2D annotations, we automatically reconstruct 3D plane equations for the structural elements and their spatial extent in the scene, and connect adjacent elements at the appropriate contact edges. We annotate and publicly release 2266 3D room layouts on the RealEstate10k dataset, containing YouTube videos. We demonstrate the high quality of these 3D layouts annotations with extensive experiments.

Abstract: Event cameras triggered a paradigm shift in the computer vision community delineated by their asynchronous nature, low latency, and high dynamic range. Calibration of event cameras is always essential to account for the sensor intrinsic parameters and for 3D perception. However, conventional image-based calibration techniques are not applicable due to the asynchronous, binary output of the sensor. The current standard for calibrating event cameras relies on either blinking patterns or event-based image reconstruction algorithms. These approaches are difficult to deploy in factory settings and are affected by noise and artifacts degrading the calibration performance. To bridge these limitations, we present E-Calib, a novel, fast, robust, and accurate calibration toolbox for event cameras utilizing the asymmetric circle grid, for its robustness to out-of-focus scenes. The proposed method is tested in a variety of rigorous experiments for different event camera models, on circle grids with different geometric properties, and under challenging illumination conditions. The results show that our approach outperforms the state-of-the-art in detection success rate, reprojection error, and estimation accuracy of extrinsic parameters.

Abstract: Due to the limited scale and quality of video-text training corpus, most vision-language foundation models employ image-text datasets for pretraining and primarily focus on modeling visually semantic representations while disregarding temporal semantic representations and correlations. To address this issue, we propose COSA, a COncatenated SAmple pretrained vision-language foundation model. COSA jointly models visual contents and event-level temporal cues using only image-text corpora. We achieve this by sequentially concatenating multiple image-text pairs as inputs for pretraining. This transformation effectively converts existing image-text corpora into a pseudo long-form video-paragraph corpus, enabling richer scene transformations and explicit event-description correspondence. Extensive experiments demonstrate that COSA consistently improves performance across a broad range of downstream tasks, including long-form/short-form video-text tasks and image-text tasks such as retrieval, captioning, and question answering. Notably, COSA achieves state-of-the-art results on various competitive benchmarks. Code and model are released at https://github.com/TXH-mercury/COSA.

Abstract: Poster layout is a crucial aspect of poster design. Prior methods primarily focus on the correlation between visual content and graphic elements. However, a pleasant layout should also consider the relationship between visual and textual contents and the relationship between elements. In this study, we introduce a relation-aware diffusion model for poster layout generation that incorporates these two relationships in the generation process. Firstly, we devise a visual-textual relation-aware module that aligns the visual and textual representations across modalities, thereby enhancing the layout's efficacy in conveying textual information. Subsequently, we propose a geometry relation-aware module that learns the geometry relationship between elements by comprehensively considering contextual information. Additionally, the proposed method can generate diverse layouts based on user constraints. To advance research in this field, we have constructed a poster layout dataset named CGL-Dataset V2. Our proposed method outperforms state-of-the-art methods on CGL-Dataset V2. The data and code will be available at https://github.com/liuan0803/RADM.

Abstract: To overcome the domain gap between synthetic and real-world datasets, unsupervised domain adaptation methods have been proposed for semantic segmentation. Majority of the previous approaches have attempted to reduce the gap either at the pixel or feature level, disregarding the fact that the two components interact positively. To address this, we present CONtrastive FEaTure and pIxel alignment (CONFETI) for bridging the domain gap at both the pixel and feature levels using a unique contrastive formulation. We introduce well-estimated prototypes by including category-wise cross-domain information to link the two alignments: the pixel-level alignment is achieved using the jointly trained style transfer module with the prototypical semantic consistency, while the feature-level alignment is enforced to cross-domain features with the \textbf{pixel-to-prototype contrast}. Our extensive experiments demonstrate that our method outperforms existing state-of-the-art methods using DeepLabV2. Our code is available at https://github.com/cxa9264/CONFETI

Abstract: Recent advances in neural reconstruction enable high-quality 3D object reconstruction from casually captured image collections. Current techniques mostly analyze their progress on relatively simple image collections where Structure-from-Motion (SfM) techniques can provide ground-truth (GT) camera poses. We note that SfM techniques tend to fail on in-the-wild image collections such as image search results with varying backgrounds and illuminations. To enable systematic research progress on 3D reconstruction from casual image captures, we propose NAVI: a new dataset of category-agnostic image collections of objects with high-quality 3D scans along with per-image 2D-3D alignments providing near-perfect GT camera parameters. These 2D-3D alignments allow us to extract accurate derivative annotations such as dense pixel correspondences, depth and segmentation maps. We demonstrate the use of NAVI image collections on different problem settings and show that NAVI enables more thorough evaluations that were not possible with existing datasets. We believe NAVI is beneficial for systematic research progress on 3D reconstruction and correspondence estimation. Project page: https://navidataset.github.io

Abstract: In this technical report, we present our solution, named UniOCC, for the Vision-Centric 3D occupancy prediction track in the nuScenes Open Dataset Challenge at CVPR 2023. Existing methods for occupancy prediction primarily focus on optimizing projected features on 3D volume space using 3D occupancy labels. However, the generation process of these labels is complex and expensive (relying on 3D semantic annotations), and limited by voxel resolution, they cannot provide fine-grained spatial semantics. To address this limitation, we propose a novel Unifying Occupancy (UniOcc) prediction method, explicitly imposing spatial geometry constraint and complementing fine-grained semantic supervision through volume ray rendering. Our method significantly enhances model performance and demonstrates promising potential in reducing human annotation costs. Given the laborious nature of annotating 3D occupancy, we further introduce a Depth-aware Teacher Student (DTS) framework to enhance prediction accuracy using unlabeled data. Our solution achieves 51.27\% mIoU on the official leaderboard with single model, placing 3rd in this challenge.

Abstract: Adversarial attacks in the physical world, particularly patch attacks, pose significant threats to the robustness and reliability of deep learning models. Developing reliable defenses against patch attacks is crucial for real-world applications, yet current research in this area is severely lacking. In this paper, we propose DIFFender, a novel defense method that leverages the pre-trained diffusion model to perform both localization and defense against potential adversarial patch attacks. DIFFender is designed as a pipeline consisting of two main stages: patch localization and restoration. In the localization stage, we exploit the intriguing properties of a diffusion model to effectively identify the locations of adversarial patches. In the restoration stage, we employ a text-guided diffusion model to eliminate adversarial regions in the image while preserving the integrity of the visual content. Additionally, we design a few-shot prompt-tuning algorithm to facilitate simple and efficient tuning, enabling the learned representations to easily transfer to downstream tasks, which optimize two stages jointly. We conduct extensive experiments on image classification and face recognition to demonstrate that DIFFender exhibits superior robustness under strong adaptive attacks and generalizes well across various scenarios, diverse classifiers, and multiple attack methods.

Abstract: We propose a novel loss function for imbalanced classification. LDAM loss, which minimizes a margin-based generalization bound, is widely utilized for class-imbalanced image classification. Although, by using LDAM loss, it is possible to obtain large margins for the minority classes and small margins for the majority classes, the relevance to a large margin, which is included in the original softmax cross entropy loss, is not be clarified yet. In this study, we reconvert the formula of LDAM loss using the concept of the large margin softmax cross entropy loss based on the softplus function and confirm that LDAM loss includes a wider large margin than softmax cross entropy loss. Furthermore, we propose a novel Enlarged Large Margin (ELM) loss, which can further widen the large margin of LDAM loss. ELM loss utilizes the large margin for the maximum logit of the incorrect class in addition to the basic margin used in LDAM loss. Through experiments conducted on imbalanced CIFAR datasets and large-scale datasets with long-tailed distribution, we confirmed that classification accuracy was much improved compared with LDAM loss and conventional losses for imbalanced classification.

Abstract: This paper presents a novel object detector called DEYOv2, an improved version of the first-generation DEYO (DETR with YOLO) model. DEYOv2, similar to its predecessor, DEYOv2 employs a progressive reasoning approach to accelerate model training and enhance performance. The study delves into the limitations of one-to-one matching in optimization and proposes solutions to effectively address the issue, such as Rank Feature and Greedy Matching. This approach enables the third stage of DEYOv2 to maximize information acquisition from the first and second stages without needing NMS, achieving end-to-end optimization. By combining dense queries, sparse queries, one-to-many matching, and one-to-one matching, DEYOv2 leverages the advantages of each method. It outperforms all existing query-based end-to-end detectors under the same settings. When using ResNet-50 as the backbone and multi-scale features on the COCO dataset, DEYOv2 achieves 51.1 AP and 51.8 AP in 12 and 24 epochs, respectively. Compared to the end-to-end model DINO, DEYOv2 provides significant performance gains of 2.1 AP and 1.4 AP in the two epoch settings. To the best of our knowledge, DEYOv2 is the first fully end-to-end object detector that combines the respective strengths of classical detectors and query-based detectors.

Abstract: This report presents ReLER submission to two tracks in the Ego4D Episodic Memory Benchmark in CVPR 2023, including Natural Language Queries and Moment Queries. This solution inherits from our proposed Action Sensitivity Learning framework (ASL) to better capture discrepant information of frames. Further, we incorporate a series of stronger video features and fusion strategies. Our method achieves an average mAP of 29.34, ranking 1st in Moment Queries Challenge, and garners 19.79 mean R1, ranking 2nd in Natural Language Queries Challenge. Our code will be released.

Abstract: Humans navigate in their environment by learning a mental model of the world through passive observation and active interaction. Their world model allows them to anticipate what might happen next and act accordingly with respect to an underlying objective. Such world models hold strong promises for planning in complex environments like in autonomous driving. A human driver, or a self-driving system, perceives their surroundings with their eyes or their cameras. They infer an internal representation of the world which should: (i) have spatial memory (e.g. occlusions), (ii) fill partially observable or noisy inputs (e.g. when blinded by sunlight), and (iii) be able to reason about unobservable events probabilistically (e.g. predict different possible futures). They are embodied intelligent agents that can predict, plan, and act in the physical world through their world model. In this thesis we present a general framework to train a world model and a policy, parameterised by deep neural networks, from camera observations and expert demonstrations. We leverage important computer vision concepts such as geometry, semantics, and motion to scale world models to complex urban driving scenes. First, we propose a model that predicts important quantities in computer vision: depth, semantic segmentation, and optical flow. We then use 3D geometry as an inductive bias to operate in the bird's-eye view space. We present for the first time a model that can predict probabilistic future trajectories of dynamic agents in bird's-eye view from 360{\deg} surround monocular cameras only. Finally, we demonstrate the benefits of learning a world model in closed-loop driving. Our model can jointly predict static scene, dynamic scene, and ego-behaviour in an urban driving environment.

Abstract: Score-matching and diffusion models have emerged as state-of-the-art generative models for both conditional and unconditional generation. Classifier-guided diffusion models are created by training a classifier on samples obtained from the forward-diffusion process (i.e., from data to noise). In this paper, we propose denoising-assisted (DA) classifiers wherein the diffusion classifier is trained using both noisy and denoised examples as simultaneous inputs to the model. We differentiate between denoising-assisted (DA) classifiers and noisy classifiers, which are diffusion classifiers that are only trained on noisy examples. Our experiments on Cifar10 and Imagenet show that DA-classifiers improve over noisy classifiers both quantitatively in terms of generalization to test data and qualitatively in terms of perceptually-aligned classifier-gradients and generative modeling metrics. Finally, we describe a semi-supervised framework for training diffusion classifiers and our experiments, that also include positive-unlabeled settings, demonstrate improved generalization of DA-classifiers over noisy classifiers.

Abstract: Cracks provide an essential indicator of infrastructure performance degradation, and achieving high-precision pixel-level crack segmentation is an issue of concern. Unlike the common research paradigms that adopt novel artificial intelligence (AI) methods directly, this paper examines the inherent characteristics of cracks so as to introduce boundary features into crack identification and then builds a boundary guidance crack segmentation model (BGCrack) with targeted structures and modules, including a high frequency module, global information modeling module, joint optimization module, etc. Extensive experimental results verify the feasibility of the proposed designs and the effectiveness of the edge information in improving segmentation results. In addition, considering that notable open-source datasets mainly consist of asphalt pavement cracks because of ease of access, there is no standard and widely recognized dataset yet for steel structures, one of the primary structural forms in civil infrastructure. This paper provides a steel crack dataset that establishes a unified and fair benchmark for the identification of steel cracks.

Abstract: Food image segmentation is an important task that has ubiquitous applications, such as estimating the nutritional value of a plate of food. Although machine learning models have been used for segmentation in this domain, food images pose several challenges. One challenge is that food items can overlap and mix, making them difficult to distinguish. Another challenge is the degree of inter-class similarity and intra-class variability, which is caused by the varying preparation methods and dishes a food item may be served in. Additionally, class imbalance is an inevitable issue in food datasets. To address these issues, two models are trained and compared, one based on convolutional neural networks and the other on Bidirectional Encoder representation for Image Transformers (BEiT). The models are trained and valuated using the FoodSeg103 dataset, which is identified as a robust benchmark for food image segmentation. The BEiT model outperforms the previous state-of-the-art model by achieving a mean intersection over union of 49.4 on FoodSeg103. This study provides insights into transfering knowledge using convolution and Transformer-based approaches in the food image domain.

Abstract: We propose Encyclopedic-VQA, a large scale visual question answering (VQA) dataset featuring visual questions about detailed properties of fine-grained categories and instances. It contains 221k unique question+answer pairs each matched with (up to) 5 images, resulting in a total of 1M VQA samples. Moreover, our dataset comes with a controlled knowledge base derived from Wikipedia, marking the evidence to support each answer. Empirically, we show that our dataset poses a hard challenge for large vision+language models as they perform poorly on our dataset: PaLI [14] is state-of-the-art on OK-VQA [37], yet it only achieves 13.0% accuracy on our dataset. Moreover, we experimentally show that progress on answering our encyclopedic questions can be achieved by augmenting large models with a mechanism that retrieves relevant information from the knowledge base. An oracle experiment with perfect retrieval achieves 87.0% accuracy on the single-hop portion of our dataset, and an automatic retrieval-augmented prototype yields 48.8%. We believe that our dataset enables future research on retrieval-augmented vision+language models.

Abstract: In this paper, we propose a novel text promptable surgical instrument segmentation approach to overcome challenges associated with diversity and differentiation of surgical instruments in minimally invasive surgeries. We redefine the task as text promptable, thereby enabling a more nuanced comprehension of surgical instruments and adaptability to new instrument types. Inspired by recent advancements in vision-language models, we leverage pretrained image and text encoders as our model backbone and design a text promptable mask decoder consisting of attention- and convolution-based prompting schemes for surgical instrument segmentation prediction. Our model leverages multiple text prompts for each surgical instrument through a new mixture of prompts mechanism, resulting in enhanced segmentation performance. Additionally, we introduce a hard instrument area reinforcement module to improve image feature comprehension and segmentation precision. Extensive experiments on EndoVis2017 and EndoVis2018 datasets demonstrate our model's superior performance and promising generalization capability. To our knowledge, this is the first implementation of a promptable approach to surgical instrument segmentation, offering significant potential for practical application in the field of robotic-assisted surgery.

Abstract: Fairness in machine learning is important for societal well-being, but limited public datasets hinder its progress. Currently, no dedicated public medical datasets with imaging data for fairness learning are available, though minority groups suffer from more health issues. To address this gap, we introduce Harvard Glaucoma Fairness (Harvard-GF), a retinal nerve disease dataset with both 2D and 3D imaging data and balanced racial groups for glaucoma detection. Glaucoma is the leading cause of irreversible blindness globally with Blacks having doubled glaucoma prevalence than other races. We also propose a fair identity normalization (FIN) approach to equalize the feature importance between different identity groups. Our FIN approach is compared with various the-state-of-the-arts fairness learning methods with superior performance in both racial and gender fairness tasks with 2D and 3D imaging data, which demonstrate the utilities of our dataset Harvard-GF for fairness learning. To facilitate fairness comparisons between different models, we propose an equity-scaled performance measure, which can be flexibly used to compare all kinds of performance metrics in the context of fairness. The dataset and code are publicly accessible via https://doi.org/10.7910/DVN/A4XMO1 and https://github.com/luoyan407/Harvard-GF, respectively.

Abstract: Large Vision-Language Models (LVLMs) have recently played a dominant role in multimodal vision-language learning. Despite the great success, it lacks a holistic evaluation of their efficacy. This paper presents a comprehensive evaluation of publicly available large multimodal models by building a LVLM evaluation Hub (LVLM-eHub). Our LVLM-eHub consists of $8$ representative LVLMs such as InstructBLIP and MiniGPT-4, which are thoroughly evaluated by a quantitative capability evaluation and an online arena platform. The former evaluates $6$ categories of multimodal capabilities of LVLMs such as visual question answering and embodied artificial intelligence on $47$ standard text-related visual benchmarks, while the latter provides the user-level evaluation of LVLMs in an open-world question-answering scenario. The study reveals several innovative findings. First, instruction-tuned LVLM with massive in-domain data such as InstructBLIP heavily overfits many existing tasks, generalizing poorly in the open-world scenario. Second, instruction-tuned LVLM with moderate instruction-following data may result in object hallucination issues (i.e., generate objects that are inconsistent with target images in the descriptions). It either makes the current evaluation metric such as CIDEr for image captioning ineffective or generates wrong answers. Third, employing a multi-turn reasoning evaluation framework can mitigate the issue of object hallucination, shedding light on developing an effective pipeline for LVLM evaluation. The findings provide a foundational framework for the conception and assessment of innovative strategies aimed at enhancing zero-shot multimodal techniques. Our LVLM-eHub will be available at https://github.com/OpenGVLab/Multi-Modality-Arena

Abstract: Intelligent Transportation Systems (ITS) allow a drastic expansion of the visibility range and decrease occlusions for autonomous driving. To obtain accurate detections, detailed labeled sensor data for training is required. Unfortunately, high-quality 3D labels of LiDAR point clouds from the infrastructure perspective of an intersection are still rare. Therefore, we provide the A9 Intersection Dataset, which consists of labeled LiDAR point clouds and synchronized camera images. Here, we recorded the sensor output from two roadside cameras and LiDARs mounted on intersection gantry bridges. The point clouds were labeled in 3D by experienced annotators. Furthermore, we provide calibration data between all sensors, which allow the projection of the 3D labels into the camera images and an accurate data fusion. Our dataset consists of 4.8k images and point clouds with more than 57.4k manually labeled 3D boxes. With ten object classes, it has a high diversity of road users in complex driving maneuvers, such as left and right turns, overtaking, and U-turns. In experiments, we provided multiple baselines for the perception tasks. Overall, our dataset is a valuable contribution to the scientific community to perform complex 3D camera-LiDAR roadside perception tasks. Find data, code, and more information at https://a9-dataset.com.

Abstract: This technical report outlines our submission to the zero-shot track of the Visual Anomaly and Novelty Detection (VAND) 2023 Challenge. Building on the performance of the WINCLIP framework, we aim to enhance the system's localization capabilities by integrating zero-shot segmentation models. In addition, we perform foreground instance segmentation which enables the model to focus on the relevant parts of the image, thus allowing the models to better identify small or subtle deviations. Our pipeline requires no external data or information, allowing for it to be directly applied to new datasets. Our team (Variance Vigilance Vanguard) ranked third in the zero-shot track of the VAND challenge, and achieve an average F1-max score of 81.5/24.2 at a sample/pixel level on the VisA dataset.

Abstract: This paper presents a novel free-hand sketch synthesis approach addressing explicit abstraction control in class-conditional and photo-to-sketch synthesis. Abstraction is a vital aspect of sketches, as it defines the fundamental distinction between a sketch and an image. Previous works relied on implicit control to achieve different levels of abstraction, leading to inaccurate control and synthesized sketches deviating from human sketches. To resolve this challenge, we propose two novel abstraction control mechanisms, state embeddings and the stroke token, integrated into a transformer-based latent diffusion model (LDM). These mechanisms explicitly provide the required amount of points or strokes to the model, enabling accurate point-level and stroke-level control in synthesized sketches while preserving recognizability. Outperforming state-of-the-art approaches, our method effectively generates diverse, non-rigid and human-like sketches. The proposed approach enables coherent sketch synthesis and excels in representing human habits with desired abstraction levels, highlighting the potential of sketch synthesis for real-world applications.

Abstract: Due to the increase in computational resources and accessibility of data, an increase in large, deep learning models trained on copious amounts of data using self-supervised or semi-supervised learning have emerged. These "foundation" models are often adapted to a variety of downstream tasks like classification, object detection, and segmentation with little-to-no training on the target dataset. In this work, we perform a robustness analysis of Visual Foundation Models (VFMs) for segmentation tasks and compare them to supervised models of smaller scale. We focus on robustness against real-world distribution shift perturbations.We benchmark four state-of-the-art segmentation architectures using 2 different datasets, COCO and ADE20K, with 17 different perturbations with 5 severity levels each. We find interesting insights that include (1) VFMs are not robust to compression-based corruptions, (2) while the selected VFMs do not significantly outperform or exhibit more robustness compared to non-VFM models, they remain competitively robust in zero-shot evaluations, particularly when non-VFM are under supervision and (3) selected VFMs demonstrate greater resilience to specific categories of objects, likely due to their open-vocabulary training paradigm, a feature that non-VFM models typically lack. We posit that the suggested robustness evaluation introduces new requirements for foundational models, thus sparking further research to enhance their performance.

Abstract: In few-shot recognition, a classifier that has been trained on one set of classes is required to rapidly adapt and generalize to a disjoint, novel set of classes. To that end, recent studies have shown the efficacy of fine-tuning with carefully crafted adaptation architectures. However this raises the question of: How can one design the optimal adaptation strategy? In this paper, we study this question through the lens of neural architecture search (NAS). Given a pre-trained neural network, our algorithm discovers the optimal arrangement of adapters, which layers to keep frozen and which to fine-tune. We demonstrate the generality of our NAS method by applying it to both residual networks and vision transformers and report state-of-the-art performance on Meta-Dataset and Meta-Album.

Abstract: Radar is a key component of the suite of perception sensors used for safe and reliable navigation of autonomous vehicles. Its unique capabilities include high-resolution velocity imaging, detection of agents in occlusion and over long ranges, and robust performance in adverse weather conditions. However, the usage of radar data presents some challenges: it is characterized by low resolution, sparsity, clutter, high uncertainty, and lack of good datasets. These challenges have limited radar deep learning research. As a result, current radar models are often influenced by lidar and vision models, which are focused on optical features that are relatively weak in radar data, thus resulting in under-utilization of radar's capabilities and diminishing its contribution to autonomous perception. This review seeks to encourage further deep learning research on autonomous radar data by 1) identifying key research themes, and 2) offering a comprehensive overview of current opportunities and challenges in the field. Topics covered include early and late fusion, occupancy flow estimation, uncertainty modeling, and multipath detection. The paper also discusses radar fundamentals and data representation, presents a curated list of recent radar datasets, and reviews state-of-the-art lidar and vision models relevant for radar research. For a summary of the paper and more results, visit the website: autonomous-radars.github.io.

Abstract: We propose an efficient approach to train large diffusion models with masked transformers. While masked transformers have been extensively explored for representation learning, their application to generative learning is less explored in the vision domain. Our work is the first to exploit masked training to reduce the training cost of diffusion models significantly. Specifically, we randomly mask out a high proportion (\emph{e.g.}, 50\%) of patches in diffused input images during training. For masked training, we introduce an asymmetric encoder-decoder architecture consisting of a transformer encoder that operates only on unmasked patches and a lightweight transformer decoder on full patches. To promote a long-range understanding of full patches, we add an auxiliary task of reconstructing masked patches to the denoising score matching objective that learns the score of unmasked patches. Experiments on ImageNet-256$\times$256 show that our approach achieves the same performance as the state-of-the-art Diffusion Transformer (DiT) model, using only 31\% of its original training time. Thus, our method allows for efficient training of diffusion models without sacrificing the generative performance.

Abstract: We introduce Infinigen, a procedural generator of photorealistic 3D scenes of the natural world. Infinigen is entirely procedural: every asset, from shape to texture, is generated from scratch via randomized mathematical rules, using no external source and allowing infinite variation and composition. Infinigen offers broad coverage of objects and scenes in the natural world including plants, animals, terrains, and natural phenomena such as fire, cloud, rain, and snow. Infinigen can be used to generate unlimited, diverse training data for a wide range of computer vision tasks including object detection, semantic segmentation, optical flow, and 3D reconstruction. We expect Infinigen to be a useful resource for computer vision research and beyond. Please visit https://infinigen.org for videos, code and pre-generated data.

Abstract: The variety of objects in the real world is nearly unlimited and is thus impossible to capture using models trained on a fixed set of categories. As a result, in recent years, open-vocabulary methods have attracted the interest of the community. This paper proposes a new method for zero-shot open-vocabulary segmentation. Prior work largely relies on contrastive training using image-text pairs, leveraging grouping mechanisms to learn image features that are both aligned with language and well-localised. This however can introduce ambiguity as the visual appearance of images with similar captions often varies. Instead, we leverage the generative properties of large-scale text-to-image diffusion models to sample a set of support images for a given textual category. This provides a distribution of appearances for a given text circumventing the ambiguity problem. We further propose a mechanism that considers the contextual background of the sampled images to better localise objects and segment the background directly. We show that our method can be used to ground several existing pre-trained self-supervised feature extractors in natural language and provide explainable predictions by mapping back to regions in the support set. Our proposal is training-free, relying on pre-trained components only, yet, shows strong performance on a range of open-vocabulary segmentation benchmarks, obtaining a lead of more than 10% on the Pascal VOC benchmark.

Abstract: In this study, we address local photo enhancement to improve the aesthetic quality of an input image by applying different effects to different regions. Existing photo enhancement methods are either not content-aware or not local; therefore, we propose a crowd-powered local enhancement method for content-aware local enhancement, which is achieved by asking crowd workers to locally optimize parameters for image editing functions. To make it easier to locally optimize the parameters, we propose an active learning based local filter. The parameters need to be determined at only a few key pixels selected by an active learning method, and the parameters at the other pixels are automatically predicted using a regression model. The parameters at the selected key pixels are independently optimized, breaking down the optimization problem into a sequence of single-slider adjustments. Our experiments show that the proposed filter outperforms existing filters, and our enhanced results are more visually pleasing than the results by the existing enhancement methods. Our source code and results are available at https://github.com/satoshi-kosugi/crowd-powered.

Abstract: Reconstructing and relighting objects and scenes under varying lighting conditions is challenging: existing neural rendering methods often cannot handle the complex interactions between materials and light. Incorporating pre-computed radiance transfer techniques enables global illumination, but still struggles with materials with subsurface scattering effects. We propose a novel framework for learning the radiance transfer field via volume rendering and utilizing various appearance cues to refine geometry end-to-end. This framework extends relighting and reconstruction capabilities to handle a wider range of materials in a data-driven fashion. The resulting models produce plausible rendering results in existing and novel conditions. We will release our code and a novel light stage dataset of objects with subsurface scattering effects publicly available.

Abstract: Visual Query Localization on long-form egocentric videos requires spatio-temporal search and localization of visually specified objects and is vital to build episodic memory systems. Prior work develops complex multi-stage pipelines that leverage well-established object detection and tracking methods to perform VQL. However, each stage is independently trained and the complexity of the pipeline results in slow inference speeds. We propose VQLoC, a novel single-stage VQL framework that is end-to-end trainable. Our key idea is to first build a holistic understanding of the query-video relationship and then perform spatio-temporal localization in a single shot manner. Specifically, we establish the query-video relationship by jointly considering query-to-frame correspondences between the query and each video frame and frame-to-frame correspondences between nearby video frames. Our experiments demonstrate that our approach outperforms prior VQL methods by 20% accuracy while obtaining a 10x improvement in inference speed. VQLoC is also the top entry on the Ego4D VQ2D challenge leaderboard. Project page: https://hwjiang1510.github.io/VQLoC/

Abstract: We propose a method to recommend music for an input video while allowing a user to guide music selection with free-form natural language. A key challenge of this problem setting is that existing music video datasets provide the needed (video, music) training pairs, but lack text descriptions of the music. This work addresses this challenge with the following three contributions. First, we propose a text-synthesis approach that relies on an analogy-based prompting procedure to generate natural language music descriptions from a large-scale language model (BLOOM-176B) given pre-trained music tagger outputs and a small number of human text descriptions. Second, we use these synthesized music descriptions to train a new trimodal model, which fuses text and video input representations to query music samples. For training, we introduce a text dropout regularization mechanism which we show is critical to model performance. Our model design allows for the retrieved music audio to agree with the two input modalities by matching visual style depicted in the video and musical genre, mood, or instrumentation described in the natural language query. Third, to evaluate our approach, we collect a testing dataset for our problem by annotating a subset of 4k clips from the YT8M-MusicVideo dataset with natural language music descriptions which we make publicly available. We show that our approach can match or exceed the performance of prior methods on video-to-music retrieval while significantly improving retrieval accuracy when using text guidance.

Abstract: We present DreamHuman, a method to generate realistic animatable 3D human avatar models solely from textual descriptions. Recent text-to-3D methods have made considerable strides in generation, but are still lacking in important aspects. Control and often spatial resolution remain limited, existing methods produce fixed rather than animated 3D human models, and anthropometric consistency for complex structures like people remains a challenge. DreamHuman connects large text-to-image synthesis models, neural radiance fields, and statistical human body models in a novel modeling and optimization framework. This makes it possible to generate dynamic 3D human avatars with high-quality textures and learned, instance-specific, surface deformations. We demonstrate that our method is capable to generate a wide variety of animatable, realistic 3D human models from text. Our 3D models have diverse appearance, clothing, skin tones and body shapes, and significantly outperform both generic text-to-3D approaches and previous text-based 3D avatar generators in visual fidelity. For more results and animations please check our website at https://dream-human.github.io.

Abstract: Arbitrary Style Transfer (AST) aims to transform images by adopting the style from any selected artwork. Nonetheless, the need to accommodate diverse and subjective user preferences poses a significant challenge. While some users wish to preserve distinct content structures, others might favor a more pronounced stylization. Despite advances in feed-forward AST methods, their limited customizability hinders their practical application. We propose a new approach, ArtFusion, which provides a flexible balance between content and style. In contrast to traditional methods reliant on biased similarity losses, ArtFusion utilizes our innovative Dual Conditional Latent Diffusion Probabilistic Models (Dual-cLDM). This approach mitigates repetitive patterns and enhances subtle artistic aspects like brush strokes and genre-specific features. Despite the promising results of conditional diffusion probabilistic models (cDM) in various generative tasks, their introduction to style transfer is challenging due to the requirement for paired training data. ArtFusion successfully navigates this issue, offering more practical and controllable stylization. A key element of our approach involves using a single image for both content and style during model training, all the while maintaining effective stylization during inference. ArtFusion outperforms existing approaches on outstanding controllability and faithful presentation of artistic details, providing evidence of its superior style transfer capabilities. Furthermore, the Dual-cLDM utilized in ArtFusion carries the potential for a variety of complex multi-condition generative tasks, thus greatly broadening the impact of our research.

Abstract: Human perception of surroundings is often guided by the various poses present within the environment. Many computer vision tasks, such as human action recognition and robot imitation learning, rely on pose-based entities like human skeletons or robotic arms. However, conventional Vision Transformer (ViT) models uniformly process all patches, neglecting valuable pose priors in input videos. We argue that incorporating poses into RGB data is advantageous for learning fine-grained and viewpoint-agnostic representations. Consequently, we introduce two strategies for learning pose-aware representations in ViTs. The first method, called Pose-aware Attention Block (PAAB), is a plug-and-play ViT block that performs localized attention on pose regions within videos. The second method, dubbed Pose-Aware Auxiliary Task (PAAT), presents an auxiliary pose prediction task optimized jointly with the primary ViT task. Although their functionalities differ, both methods succeed in learning pose-aware representations, enhancing performance in multiple diverse downstream tasks. Our experiments, conducted across seven datasets, reveal the efficacy of both pose-aware methods on three video analysis tasks, with PAAT holding a slight edge over PAAB. Both PAAT and PAAB surpass their respective backbone Transformers by up to 9.8% in real-world action recognition and 21.8% in multi-view robotic video alignment. Code is available at https://github.com/dominickrei/PoseAwareVT.

Abstract: We address personalized image enhancement in this study, where we enhance input images for each user based on the user's preferred images. Previous methods apply the same preferred style to all input images (i.e., only one style for each user); in contrast to these methods, we aim to achieve content-aware personalization by applying different styles to each image considering the contents. For content-aware personalization, we make two contributions. First, we propose a method named masked style modeling, which can predict a style for an input image considering the contents by using the framework of masked language modeling. Second, to allow this model to consider the contents of images, we propose a novel training scheme where we download images from Flickr and create pseudo input and retouched image pairs using a degrading model. We conduct quantitative evaluations and a user study, and our method trained using our training scheme successfully achieves content-aware personalization; moreover, our method outperforms other previous methods in this field. Our source code is available at https://github.com/satoshi-kosugi/masked-style-modeling.

Abstract: Social interaction is a fundamental aspect of human behavior and communication. The way individuals position themselves in relation to others, also known as proxemics, conveys social cues and affects the dynamics of social interaction. We present a novel approach that learns a 3D proxemics prior of two people in close social interaction. Since collecting a large 3D dataset of interacting people is a challenge, we rely on 2D image collections where social interactions are abundant. We achieve this by reconstructing pseudo-ground truth 3D meshes of interacting people from images with an optimization approach using existing ground-truth contact maps. We then model the proxemics using a novel denoising diffusion model called BUDDI that learns the joint distribution of two people in close social interaction directly in the SMPL-X parameter space. Sampling from our generative proxemics model produces realistic 3D human interactions, which we validate through a user study. Additionally, we introduce a new optimization method that uses the diffusion prior to reconstruct two people in close proximity from a single image without any contact annotation. Our approach recovers more accurate and plausible 3D social interactions from noisy initial estimates and outperforms state-of-the-art methods. See our project site for code, data, and model: muelea.github.io/buddi.

Abstract: Recent text-to-image generative models can generate high-fidelity images from text inputs, but the quality of these generated images cannot be accurately evaluated by existing evaluation metrics. To address this issue, we introduce Human Preference Dataset v2 (HPD v2), a large-scale dataset that captures human preferences on images from a wide range of sources. HPD v2 comprises 798,090 human preference choices on 430,060 pairs of images, making it the largest dataset of its kind. The text prompts and images are deliberately collected to eliminate potential bias, which is a common issue in previous datasets. By fine-tuning CLIP on HPD v2, we obtain Human Preference Score v2 (HPS v2), a scoring model that can more accurately predict text-generated images' human preferences. Our experiments demonstrate that HPS v2 generalizes better than previous metrics across various image distributions and is responsive to algorithmic improvements of text-to-image generative models, making it a preferable evaluation metric for these models. We also investigate the design of the evaluation prompts for text-to-image generative models, to make the evaluation stable, fair and easy-to-use. Finally, we establish a benchmark for text-to-image generative models using HPS v2, which includes a set of recent text-to-image models from the academia, community and industry. The code and dataset is / will be available at https://github.com/tgxs002/HPSv2.

Abstract: Current perceptual similarity metrics operate at the level of pixels and patches. These metrics compare images in terms of their low-level colors and textures, but fail to capture mid-level similarities and differences in image layout, object pose, and semantic content. In this paper, we develop a perceptual metric that assesses images holistically. Our first step is to collect a new dataset of human similarity judgments over image pairs that are alike in diverse ways. Critical to this dataset is that judgments are nearly automatic and shared by all observers. To achieve this we use recent text-to-image models to create synthetic pairs that are perturbed along various dimensions. We observe that popular perceptual metrics fall short of explaining our new data, and we introduce a new metric, DreamSim, tuned to better align with human perception. We analyze how our metric is affected by different visual attributes, and find that it focuses heavily on foreground objects and semantic content while also being sensitive to color and layout. Notably, despite being trained on synthetic data, our metric generalizes to real images, giving strong results on retrieval and reconstruction tasks. Furthermore, our metric outperforms both prior learned metrics and recent large vision models on these tasks.

Abstract: While large text-to-image models are able to synthesize "novel" images, these images are necessarily a reflection of the training data. The problem of data attribution in such models -- which of the images in the training set are most responsible for the appearance of a given generated image -- is a difficult yet important one. As an initial step toward this problem, we evaluate attribution through "customization" methods, which tune an existing large-scale model toward a given exemplar object or style. Our key insight is that this allows us to efficiently create synthetic images that are computationally influenced by the exemplar by construction. With our new dataset of such exemplar-influenced images, we are able to evaluate various data attribution algorithms and different possible feature spaces. Furthermore, by training on our dataset, we can tune standard models, such as DINO, CLIP, and ViT, toward the attribution problem. Even though the procedure is tuned towards small exemplar sets, we show generalization to larger sets. Finally, by taking into account the inherent uncertainty of the problem, we can assign soft attribution scores over a set of training images.

Abstract: Recent advancements in vision foundation models (VFMs) have opened up new possibilities for versatile and efficient visual perception. In this work, we introduce Seal, a novel framework that harnesses VFMs for segmenting diverse automotive point cloud sequences. Seal exhibits three appealing properties: i) Scalability: VFMs are directly distilled into point clouds, eliminating the need for annotations in either 2D or 3D during pretraining. ii) Consistency: Spatial and temporal relationships are enforced at both the camera-to-LiDAR and point-to-segment stages, facilitating cross-modal representation learning. iii) Generalizability: Seal enables knowledge transfer in an off-the-shelf manner to downstream tasks involving diverse point clouds, including those from real/synthetic, low/high-resolution, large/small-scale, and clean/corrupted datasets. Extensive experiments conducted on eleven different point cloud datasets showcase the effectiveness and superiority of Seal. Notably, Seal achieves a remarkable 45.0% mIoU on nuScenes after linear probing, surpassing random initialization by 36.9% mIoU and outperforming prior arts by 6.1% mIoU. Moreover, Seal demonstrates significant performance gains over existing methods across 20 different few-shot fine-tuning tasks on all eleven tested point cloud datasets.

Abstract: Do different neural networks, trained for various vision tasks, share some common representations? In this paper, we demonstrate the existence of common features we call "Rosetta Neurons" across a range of models with different architectures, different tasks (generative and discriminative), and different types of supervision (class-supervised, text-supervised, self-supervised). We present an algorithm for mining a dictionary of Rosetta Neurons across several popular vision models: Class Supervised-ResNet50, DINO-ResNet50, DINO-ViT, MAE, CLIP-ResNet50, BigGAN, StyleGAN-2, StyleGAN-XL. Our findings suggest that certain visual concepts and structures are inherently embedded in the natural world and can be learned by different models regardless of the specific task or architecture, and without the use of semantic labels. We can visualize shared concepts directly due to generative models included in our analysis. The Rosetta Neurons facilitate model-to-model translation enabling various inversion-based manipulations, including cross-class alignments, shifting, zooming, and more, without the need for specialized training.

Abstract: We show how to build a model that allows realistic, free-viewpoint renderings of a scene under novel lighting conditions from video. Our method -- UrbanIR: Urban Scene Inverse Rendering -- computes an inverse graphics representation from the video. UrbanIR jointly infers shape, albedo, visibility, and sun and sky illumination from a single video of unbounded outdoor scenes with unknown lighting. UrbanIR uses videos from cameras mounted on cars (in contrast to many views of the same points in typical NeRF-style estimation). As a result, standard methods produce poor geometry estimates (for example, roofs), and there are numerous ''floaters''. Errors in inverse graphics inference can result in strong rendering artifacts. UrbanIR uses novel losses to control these and other sources of error. UrbanIR uses a novel loss to make very good estimates of shadow volumes in the original scene. The resulting representations facilitate controllable editing, delivering photorealistic free-viewpoint renderings of relit scenes and inserted objects. Qualitative evaluation demonstrates strong improvements over the state-of-the-art.

Abstract: The reflective nature of the human eye is an underappreciated source of information about what the world around us looks like. By imaging the eyes of a moving person, we can collect multiple views of a scene outside the camera's direct line of sight through the reflections in the eyes. In this paper, we reconstruct a 3D scene beyond the camera's line of sight using portrait images containing eye reflections. This task is challenging due to 1) the difficulty of accurately estimating eye poses and 2) the entangled appearance of the eye iris and the scene reflections. Our method jointly refines the cornea poses, the radiance field depicting the scene, and the observer's eye iris texture. We further propose a simple regularization prior on the iris texture pattern to improve reconstruction quality. Through various experiments on synthetic and real-world captures featuring people with varied eye colors, we demonstrate the feasibility of our approach to recover 3D scenes using eye reflections.