Computer Vision and Pattern Recognition (cs.CV)

Abstract: Although the recent rapid evolution of 3D generative neural networks greatly improves 3D shape generation, it is still not convenient for ordinary users to create 3D shapes and control the local geometry of generated shapes. To address these challenges, we propose a diffusion-based 3D generation framework -- locally attentional SDF diffusion, to model plausible 3D shapes, via 2D sketch image input. Our method is built on a two-stage diffusion model. The first stage, named occupancy-diffusion, aims to generate a low-resolution occupancy field to approximate the shape shell. The second stage, named SDF-diffusion, synthesizes a high-resolution signed distance field within the occupied voxels determined by the first stage to extract fine geometry. Our model is empowered by a novel view-aware local attention mechanism for image-conditioned shape generation, which takes advantage of 2D image patch features to guide 3D voxel feature learning, greatly improving local controllability and model generalizability. Through extensive experiments in sketch-conditioned and category-conditioned 3D shape generation tasks, we validate and demonstrate the ability of our method to provide plausible and diverse 3D shapes, as well as its superior controllability and generalizability over existing work. Our code and trained models are available at https://zhengxinyang.github.io/projects/LAS-Diffusion.html

Abstract: We introduce a new problem in unsupervised domain adaptation, termed as Generalized Universal Domain Adaptation (GUDA), which aims to achieve precise prediction of all target labels including unknown categories. GUDA bridges the gap between label distribution shift-based and label space mismatch-based variants, essentially categorizing them as a unified problem, guiding to a comprehensive framework for thoroughly solving all the variants. The key challenge of GUDA is developing and identifying novel target categories while estimating the target label distribution. To address this problem, we take advantage of the powerful exploration capability of generative flow networks and propose an active domain adaptation algorithm named GFlowDA, which selects diverse samples with probabilities proportional to a reward function. To enhance the exploration capability and effectively perceive the target label distribution, we tailor the states and rewards, and introduce an efficient solution for parent exploration and state transition. We also propose a training paradigm for GUDA called Generalized Universal Adversarial Network (GUAN), which involves collaborative optimization between GUAN and GFlowNet. Theoretical analysis highlights the importance of exploration, and extensive experiments on benchmark datasets demonstrate the superiority of GFlowDA.

Abstract: In this paper, we introduce semi-supervised video object segmentation (VOS) to panoptic wild scenes and present a large-scale benchmark as well as a baseline method for it. Previous benchmarks for VOS with sparse annotations are not sufficient to train or evaluate a model that needs to process all possible objects in real-world scenarios. Our new benchmark (VIPOSeg) contains exhaustive object annotations and covers various real-world object categories which are carefully divided into subsets of thing/stuff and seen/unseen classes for comprehensive evaluation. Considering the challenges in panoptic VOS, we propose a strong baseline method named panoptic object association with transformers (PAOT), which uses panoptic identification to associate objects with a pyramid architecture on multiple scales. Experimental results show that VIPOSeg can not only boost the performance of VOS models by panoptic training but also evaluate them comprehensively in panoptic scenes. Previous methods for classic VOS still need to improve in performance and efficiency when dealing with panoptic scenes, while our PAOT achieves SOTA performance with good efficiency on VIPOSeg and previous VOS benchmarks. PAOT also ranks 1st in the VOT2022 challenge. Our dataset is available at https://github.com/yoxu515/VIPOSeg-Benchmark.

Abstract: Cross-modal contrastive learning in vision language pretraining (VLP) faces the challenge of (partial) false negatives. In this paper, we study this problem from the perspective of Mutual Information (MI) optimization. It is common sense that InfoNCE loss used in contrastive learning will maximize the lower bound of MI between anchors and their positives, while we theoretically prove that MI involving negatives also matters when noises commonly exist. Guided by a more general lower bound form for optimization, we propose a contrastive learning strategy regulated by progressively refined cross-modal similarity, to more accurately optimize MI between an image/text anchor and its negative texts/images instead of improperly minimizing it. Our method performs competitively on four downstream cross-modal tasks and systematically balances the beneficial and harmful effects of (partial) false negative samples under theoretical guidance.

Abstract: Text-to-Image (T2I) ReID has attracted a lot of attention in the recent past. CUHK-PEDES, RSTPReid and ICFG-PEDES are the three available benchmarks to evaluate T2I ReID methods. RSTPReid and ICFG-PEDES comprise of identities from MSMT17 but due to limited number of unique persons, the diversity is limited. On the other hand, CUHK-PEDES comprises of 13,003 identities but has relatively shorter text description on average. Further, these datasets are captured in a restricted environment with limited number of cameras. In order to further diversify the identities and provide dense captions, we propose a novel dataset called IIITD-20K. IIITD-20K comprises of 20,000 unique identities captured in the wild and provides a rich dataset for text-to-image ReID. With a minimum of 26 words for a description, each image is densely captioned. We further synthetically generate images and fine-grained captions using Stable-diffusion and BLIP models trained on our dataset. We perform elaborate experiments using state-of-art text-to-image ReID models and vision-language pre-trained models and present a comprehensive analysis of the dataset. Our experiments also reveal that synthetically generated data leads to a substantial performance improvement in both same dataset as well as cross dataset settings. Our dataset is available at https://bit.ly/3pkA3Rj.

Abstract: Building footprint information is an essential ingredient for 3-D reconstruction of urban models. The automatic generation of building footprints from satellite images presents a considerable challenge due to the complexity of building shapes. Recent developments in deep convolutional neural networks (DCNNs) have enabled accurate pixel-level labeling tasks. One central issue remains, which is the precise delineation of boundaries. Deep architectures generally fail to produce fine-grained segmentation with accurate boundaries due to progressive downsampling. In this work, we have proposed a end-to-end framework to overcome this issue, which uses the graph convolutional network (GCN) for building footprint extraction task. Our proposed framework outperforms state-of-the-art methods.

Abstract: It is critical for vehicles to prevent any collisions with pedestrians. Current methods for pedestrian collision prevention focus on integrating visual pedestrian detectors with Automatic Emergency Braking (AEB) systems which can trigger warnings and apply brakes as a pedestrian enters a vehicle's path. Unfortunately, pedestrian-detection-based systems can be hindered in certain situations such as night-time or when pedestrians are occluded. Our system addresses such issues using an online, map-based pedestrian detection aggregation system where common pedestrian locations are learned after repeated passes of locations. Using a carefully collected and annotated dataset in La Jolla, CA, we demonstrate the system's ability to learn pedestrian zones and generate advisory notices when a vehicle is approaching a pedestrian despite challenges like dark lighting or pedestrian occlusion. Using the number of correct advisories, false advisories, and missed advisories to define precision and recall performance metrics, we evaluate our system and discuss future positive effects with further data collection. We have made our code available at https://github.com/s7desai/ped-mapping, and a video demonstration of the CHAMP system at https://youtu.be/dxeCrS_Gpkw.

Abstract: One of the most important tasks for ensuring safe autonomous driving systems is accurately detecting road traffic lights and accurately determining how they impact the driver's actions. In various real-world driving situations, a scene may have numerous traffic lights with varying levels of relevance to the driver, and thus, distinguishing and detecting the lights that are relevant to the driver and influence the driver's actions is a critical safety task. This paper proposes a traffic light detection model which focuses on this task by first defining salient lights as the lights that affect the driver's future decisions. We then use this salience property to construct the LAVA Salient Lights Dataset, the first US traffic light dataset with an annotated salience property. Subsequently, we train a Deformable DETR object detection transformer model using Salience-Sensitive Focal Loss to emphasize stronger performance on salient traffic lights, showing that a model trained with this loss function has stronger recall than one trained without.

Abstract: Blind face restoration (BFR) is important while challenging. Prior works prefer to exploit GAN-based frameworks to tackle this task due to the balance of quality and efficiency. However, these methods suffer from poor stability and adaptability to long-tail distribution, failing to simultaneously retain source identity and restore detail. We propose DiffBFR to introduce Diffusion Probabilistic Model (DPM) for BFR to tackle the above problem, given its superiority over GAN in aspects of avoiding training collapse and generating long-tail distribution. DiffBFR utilizes a two-step design, that first restores identity information from low-quality images and then enhances texture details according to the distribution of real faces. This design is implemented with two key components: 1) Identity Restoration Module (IRM) for preserving the face details in results. Instead of denoising from pure Gaussian random distribution with LQ images as the condition during the reverse process, we propose a novel truncated sampling method which starts from LQ images with part noise added. We theoretically prove that this change shrinks the evidence lower bound of DPM and then restores more original details. With theoretical proof, two cascade conditional DPMs with different input sizes are introduced to strengthen this sampling effect and reduce training difficulty in the high-resolution image generated directly. 2) Texture Enhancement Module (TEM) for polishing the texture of the image. Here an unconditional DPM, a LQ-free model, is introduced to further force the restorations to appear realistic. We theoretically proved that this unconditional DPM trained on pure HQ images contributes to justifying the correct distribution of inference images output from IRM in pixel-level space. Truncated sampling with fractional time step is utilized to polish pixel-level textures while preserving identity information.

Abstract: Employing a dictionary can efficiently rectify the deviation between the visual prediction and the ground truth in scene text recognition methods. However, the independence of the dictionary on the visual features may lead to incorrect rectification of accurate visual predictions. In this paper, we propose a new dictionary language model leveraging the Scene Image-Text Matching(SITM) network, which avoids the drawbacks of the explicit dictionary language model: 1) the independence of the visual features; 2) noisy choice in candidates etc. The SITM network accomplishes this by using Image-Text Contrastive (ITC) Learning to match an image with its corresponding text among candidates in the inference stage. ITC is widely used in vision-language learning to pull the positive image-text pair closer in feature space. Inspired by ITC, the SITM network combines the visual features and the text features of all candidates to identify the candidate with the minimum distance in the feature space. Our lexicon method achieves better results(93.8\% accuracy) than the ordinary method results(92.1\% accuracy) on six mainstream benchmarks. Additionally, we integrate our method with ABINet and establish new state-of-the-art results on several benchmarks.

Abstract: Model compression has become the de-facto approach for optimizing the efficiency of vision models. Recently, the focus of most compression efforts has shifted to post-training scenarios due to the very high cost of large-scale pretraining. This has created the need to build compressible models from scratch, which can effectively be compressed after training. In this work, we present a sharpness-minimizing network transformation (SNT) method applied during pretraining that can create models with desirable compressibility and generalizability features. We compare our approach to a well-known sharpness-minimizing optimizer to validate its efficacy in creating a flat loss landscape. To the best of our knowledge, SNT is the first pretraining method that uses an architectural transformation to generate compression-friendly networks. We find that SNT generalizes across different compression tasks and network backbones, delivering consistent improvements over the ADAM baseline with up to 2% accuracy improvement on weight pruning and 5.4% accuracy improvement on quantization. Code to reproduce our results will be made publicly available.

Abstract: Smart home device detection is a critical aspect of human-computer interaction. However, detecting targets in indoor environments can be challenging due to interference from ambient light and background noise. In this paper, we present a new model called FSA-YOLOv5, which addresses the limitations of traditional convolutional neural networks by introducing the Transformer to learn long-range dependencies. Additionally, we propose a new attention module, the full-separation attention module, which integrates spatial and channel dimensional information to learn contextual information. To improve tiny device detection, we include a prediction head for the indoor smart home device detection task. We also release the Southeast University Indoor Smart Speaker Dataset (SUSSD) to supplement existing data samples. Through a series of experiments on SUSSD, we demonstrate that our method outperforms other methods, highlighting the effectiveness of FSA-YOLOv5.

Abstract: Long-tailed multi-label visual recognition (LTML) task is a highly challenging task due to the label co-occurrence and imbalanced data distribution. In this work, we propose a unified framework for LTML, namely prompt tuning with class-specific embedding loss (LMPT), capturing the semantic feature interactions between categories by combining text and image modality data and improving the performance synchronously on both head and tail classes. Specifically, LMPT introduces the embedding loss function with class-aware soft margin and re-weighting to learn class-specific contexts with the benefit of textual descriptions (captions), which could help establish semantic relationships between classes, especially between the head and tail classes. Furthermore, taking into account the class imbalance, the distribution-balanced loss is adopted as the classification loss function to further improve the performance on the tail classes without compromising head classes. Extensive experiments are conducted on VOC-LT and COCO-LT datasets, which demonstrates that the proposed method significantly surpasses the previous state-of-the-art methods and zero-shot CLIP in LTML. Our codes are fully available at \url{https://github.com/richard-peng-xia/LMPT}.

Abstract: Multi-baseline interferometric synthetic aperture radar (InSAR) techniques are effective approaches for retrieving the 3-D information of urban areas. In order to obtain a plausible reconstruction, it is necessary to use large-stack interferograms. Hence, these methods are commonly not appropriate for large-scale 3-D urban mapping using TanDEM-X data where only a few acquisitions are available in average for each city. This work proposes a new SAR tomographic processing framework to work with those extremely small stacks, which integrates the non-local filtering into SAR tomography inversion. The applicability of the algorithm is demonstrated using a TanDEM-X multi-baseline stack with 5 bistatic interferograms over the whole city of Munich, Germany. Systematic comparison of our result with airborne LiDAR data shows that the relative height accuracy of two third buildings is within two meters, which outperforms the TanDEM-X raw DEM. The promising performance of the proposed algorithm paved the first step towards high quality large-scale 3-D urban mapping.

Abstract: Visual Speech Recognition (VSR) is a task to predict a sentence or word from lip movements. Some works have been recently presented which use audio signals to supplement visual information. However, existing methods utilize only limited information such as phoneme-level features and soft labels of Automatic Speech Recognition (ASR) networks. In this paper, we present a Multi-Temporal Lip-Audio Memory (MTLAM) that makes the best use of audio signals to complement insufficient information of lip movements. The proposed method is mainly composed of two parts: 1) MTLAM saves multi-temporal audio features produced from short- and long-term audio signals, and the MTLAM memorizes a visual-to-audio mapping to load stored multi-temporal audio features from visual features at the inference phase. 2) We design an audio temporal model to produce multi-temporal audio features capturing the context of neighboring words. In addition, to construct effective visual-to-audio mapping, the audio temporal models can generate audio features time-aligned with visual features. Through extensive experiments, we validate the effectiveness of the MTLAM achieving state-of-the-art performances on two public VSR datasets.

Abstract: Visual localization is the task of estimating the camera pose from which a given image was taken and is central to several 3D computer vision applications. With the rapid growth in the popularity of AR/VR/MR devices and cloud-based applications, privacy issues are becoming a very important aspect of the localization process. Existing work on privacy-preserving localization aims to defend against an attacker who has access to a cloud-based service. In this paper, we show that an attacker can learn about details of a scene without any access by simply querying a localization service. The attack is based on the observation that modern visual localization algorithms are robust to variations in appearance and geometry. While this is in general a desired property, it also leads to algorithms localizing objects that are similar enough to those present in a scene. An attacker can thus query a server with a large enough set of images of objects, \eg, obtained from the Internet, and some of them will be localized. The attacker can thus learn about object placements from the camera poses returned by the service (which is the minimal information returned by such a service). In this paper, we develop a proof-of-concept version of this attack and demonstrate its practical feasibility. The attack does not place any requirements on the localization algorithm used, and thus also applies to privacy-preserving representations. Current work on privacy-preserving representations alone is thus insufficient.

Abstract: Instance-level segmentation of documents consists in assigning a class-aware and instance-aware label to each pixel of the image. It is a key step in document parsing for their understanding. In this paper, we present a unified transformer encoder-decoder architecture for en-to-end instance segmentation of complex layouts in document images. The method adapts a contrastive training with a mixed query selection for anchor initialization in the decoder. Later on, it performs a dot product between the obtained query embeddings and the pixel embedding map (coming from the encoder) for semantic reasoning. Extensive experimentation on competitive benchmarks like PubLayNet, PRIMA, Historical Japanese (HJ), and TableBank demonstrate that our model with SwinL backbone achieves better segmentation performance than the existing state-of-the-art approaches with the average precision of \textbf{93.72}, \textbf{54.39}, \textbf{84.65} and \textbf{98.04} respectively under one billion parameters. The code is made publicly available at: \href{https://github.com/ayanban011/SwinDocSegmenter}{github.com/ayanban011/SwinDocSegmenter}

Abstract: In this paper, we introduce a novel framework for the challenging problem of One-Shot Unsupervised Domain Adaptation (OSUDA), which aims to adapt to a target domain with only a single unlabeled target sample. Unlike existing approaches that rely on large labeled source and unlabeled target data, our Target-driven One-Shot UDA (TOS-UDA) approach employs a learnable augmentation strategy guided by the target sample's style to align the source distribution with the target distribution. Our method consists of three modules: an augmentation module, a style alignment module, and a classifier. Unlike existing methods, our augmentation module allows for strong transformations of the source samples, and the style of the single target sample available is exploited to guide the augmentation by ensuring perceptual similarity. Furthermore, our approach integrates augmentation with style alignment, eliminating the need for separate pre-training on additional datasets. Our method outperforms or performs comparably to existing OS-UDA methods on the Digits and DomainNet benchmarks.

Abstract: Large-scale text-to-image models have demonstrated amazing ability to synthesize diverse and high-fidelity images. However, these models are often violated by several limitations. Firstly, they require the user to provide precise and contextually relevant descriptions for the desired image modifications. Secondly, current models can impose significant changes to the original image content during the editing process. In this paper, we explore ReGeneration learning in an image-to-image Diffusion model (ReDiffuser), that preserves the content of the original image without human prompting and the requisite editing direction is automatically discovered within the text embedding space. To ensure consistent preservation of the shape during image editing, we propose cross-attention guidance based on regeneration learning. This novel approach allows for enhanced expression of the target domain features while preserving the original shape of the image. In addition, we introduce a cooperative update strategy, which allows for efficient preservation of the original shape of an image, thereby improving the quality and consistency of shape preservation throughout the editing process. Our proposed method leverages an existing pre-trained text-image diffusion model without any additional training. Extensive experiments show that the proposed method outperforms existing work in both real and synthetic image editing.

Abstract: Scale invariance of an algorithm refers to its ability to treat objects equally independently of their size. For neural networks, scale invariance is typically achieved by data augmentation. However, when presented with a scale far outside the range covered by the training set, neural networks may fail to generalize. Here, we introduce the Riesz network, a novel scale invariant neural network. Instead of standard 2d or 3d convolutions for combining spatial information, the Riesz network is based on the Riesz transform which is a scale equivariant operation. As a consequence, this network naturally generalizes to unseen or even arbitrary scales in a single forward pass. As an application example, we consider detecting and segmenting cracks in tomographic images of concrete. In this context, 'scale' refers to the crack thickness which may vary strongly even within the same sample. To prove its scale invariance, the Riesz network is trained on one fixed crack width. We then validate its performance in segmenting simulated and real tomographic images featuring a wide range of crack widths. An additional experiment is carried out on the MNIST Large Scale data set.

Abstract: Point cloud data has been extensively studied due to its compact form and flexibility in representing complex 3D structures. The ability of point cloud data to accurately capture and represent intricate 3D geometry makes it an ideal choice for a wide range of applications, including computer vision, robotics, and autonomous driving, all of which require an understanding of the underlying spatial structures. Given the challenges associated with annotating large-scale point clouds, self-supervised point cloud representation learning has attracted increasing attention in recent years. This approach aims to learn generic and useful point cloud representations from unlabeled data, circumventing the need for extensive manual annotations. In this paper, we present a comprehensive survey of self-supervised point cloud representation learning using DNNs. We begin by presenting the motivation and general trends in recent research. We then briefly introduce the commonly used datasets and evaluation metrics. Following that, we delve into an extensive exploration of self-supervised point cloud representation learning methods based on these techniques. Finally, we share our thoughts on some of the challenges and potential issues that future research in self-supervised learning for pre-training 3D point clouds may encounter.

Abstract: We present ElasticHash, a novel approach for high-quality, efficient, and large-scale semantic image similarity search. It is based on a deep hashing model to learn hash codes for fine-grained image similarity search in natural images and a two-stage method for efficiently searching binary hash codes using Elasticsearch (ES). In the first stage, a coarse search based on short hash codes is performed using multi-index hashing and ES terms lookup of neighboring hash codes. In the second stage, the list of results is re-ranked by computing the Hamming distance on long hash codes. We evaluate the retrieval performance of \textit{ElasticHash} for more than 120,000 query images on about 6.9 million database images of the OpenImages data set. The results show that our approach achieves high-quality retrieval results and low search latencies.

Abstract: In this paper, we present a novel system (denoted as Polaca) to generate poetic Chinese landscape painting with calligraphy. Unlike previous single image-to-image painting generation, Polaca takes the classic poetry as input and outputs the artistic landscape painting image with the corresponding calligraphy. It is equipped with three different modules to complete the whole piece of landscape painting artwork: the first one is a text-to-image module to generate landscape painting image, the second one is an image-to-image module to generate stylistic calligraphy image, and the third one is an image fusion module to fuse the two images into a whole piece of aesthetic artwork.

Abstract: Vision transformers (ViTs) that model an image as a sequence of partitioned patches have shown notable performance in diverse vision tasks. Because partitioning patches eliminates the image structure, to reflect the order of patches, ViTs utilize an explicit component called positional embedding. However, we claim that the use of positional embedding does not simply guarantee the order-awareness of ViT. To support this claim, we analyze the actual behavior of ViTs using an effective receptive field. We demonstrate that during training, ViT acquires an understanding of patch order from the positional embedding that is trained to be a specific pattern. Based on this observation, we propose explicitly adding a Gaussian attention bias that guides the positional embedding to have the corresponding pattern from the beginning of training. We evaluated the influence of Gaussian attention bias on the performance of ViTs in several image classification, object detection, and semantic segmentation experiments. The results showed that proposed method not only facilitates ViTs to understand images but also boosts their performance on various datasets, including ImageNet, COCO 2017, and ADE20K.

Abstract: In the modern world, one of the most severe eye infections brought on by diabetes is known as diabetic retinopathy, which will result in retinal damage, and, thus, lead to blindness. Diabetic retinopathy can be well treated with early diagnosis. Retinal fundus images of humans are used to screen for lesions in the retina. However, detecting DR in the early stages is challenging due to the minimal symptoms. Furthermore, the occurrence of diseases linked to vascular anomalies brought on by DR aids in diagnosing the condition. Nevertheless, the resources required for manually identifying the lesions are high. Similarly, training for Convolutional Neural Networks is more time-consuming. This proposed research aims to improve diabetic retinopathy diagnosis by developing an enhanced deep learning model for timely DR identification that is potentially more accurate than existing CNN-based models. The proposed model will detect various lesions from retinal images in the early stages. First, characteristics are retrieved from the retinal fundus picture and put into the EDLM for classification. For dimensionality reduction, EDLM is used. Additionally, the classification and feature extraction processes are optimized using the stochastic gradient descent optimizer. The EDLM effectiveness is assessed on the KAG GLE dataset with 3459 retinal images, and results are compared over VGG16, VGG19, RESNET18, RESNET34, and RESNET50.

Abstract: We introduce light diffusion, a novel method to improve lighting in portraits, softening harsh shadows and specular highlights while preserving overall scene illumination. Inspired by professional photographers' diffusers and scrims, our method softens lighting given only a single portrait photo. Previous portrait relighting approaches focus on changing the entire lighting environment, removing shadows (ignoring strong specular highlights), or removing shading entirely. In contrast, we propose a learning based method that allows us to control the amount of light diffusion and apply it on in-the-wild portraits. Additionally, we design a method to synthetically generate plausible external shadows with sub-surface scattering effects while conforming to the shape of the subject's face. Finally, we show how our approach can increase the robustness of higher level vision applications, such as albedo estimation, geometry estimation and semantic segmentation.

Abstract: Texture mapping as a fundamental task in 3D modeling has been well established for well-acquired aerial assets under consistent illumination, yet it remains a challenge when it is scaled to large datasets with images under varying views and illuminations. A well-performed texture mapping algorithm must be able to efficiently select views, fuse and map textures from these views to mesh models, at the same time, achieve consistent radiometry over the entire model. Existing approaches achieve efficiency either by limiting the number of images to one view per face, or simplifying global inferences to only achieve local color consistency. In this paper, we break this tie by proposing a novel and efficient texture mapping framework that allows the use of multiple views of texture per face, at the same time to achieve global color consistency. The proposed method leverages a loopy belief propagation algorithm to perform an efficient and global-level probabilistic inferences to rank candidate views per face, which enables face-level multi-view texture fusion and blending. The texture fusion algorithm, being non-parametric, brings another advantage over typical parametric post color correction methods, due to its improved robustness to non-linear illumination differences. The experiments on three different types of datasets (i.e. satellite dataset, unmanned-aerial vehicle dataset and close-range dataset) show that the proposed method has produced visually pleasant and texturally consistent results in all scenarios, with an added advantage of consuming less running time as compared to the state of the art methods, especially for large-scale dataset such as satellite-derived models.

Abstract: Semantic segmentation of multichannel images is a fundamental task for many applications. Selecting an appropriate channel combination from the original multichannel image can improve the accuracy of semantic segmentation and reduce the cost of data storage, processing and future acquisition. Existing channel selection methods typically use a reasonable selection procedure to determine a desirable channel combination, and then train a semantic segmentation network using that combination. In this study, the concept of pruning from a supernet is used for the first time to integrate the selection of channel combination and the training of a semantic segmentation network. Based on this concept, a One-Shot Task-Adaptive (OSTA) channel selection method is proposed for the semantic segmentation of multichannel images. OSTA has three stages, namely the supernet training stage, the pruning stage and the fine-tuning stage. The outcomes of six groups of experiments (L7Irish3C, L7Irish2C, L8Biome3C, L8Biome2C, RIT-18 and Semantic3D) demonstrated the effectiveness and efficiency of OSTA. OSTA achieved the highest segmentation accuracies in all tests (62.49% (mIoU), 75.40% (mIoU), 68.38% (mIoU), 87.63% (mIoU), 66.53% (mA) and 70.86% (mIoU), respectively). It even exceeded the highest accuracies of exhaustive tests (61.54% (mIoU), 74.91% (mIoU), 67.94% (mIoU), 87.32% (mIoU), 65.32% (mA) and 70.27% (mIoU), respectively), where all possible channel combinations were tested. All of this can be accomplished within a predictable and relatively efficient timeframe, ranging from 101.71% to 298.1% times the time required to train the segmentation network alone. In addition, there were interesting findings that were deemed valuable for several fields.

Abstract: The ability of deep neural networks to continually learn and adapt to a sequence of tasks has remained challenging due to catastrophic forgetting of previously learned tasks. Humans, on the other hand, have a remarkable ability to acquire, assimilate, and transfer knowledge across tasks throughout their lifetime without catastrophic forgetting. The versatility of the brain can be attributed to the rehearsal of abstract experiences through a complementary learning system. However, representation rehearsal in vision transformers lacks diversity, resulting in overfitting and consequently, performance drops significantly compared to raw image rehearsal. Therefore, we propose BiRT, a novel representation rehearsal-based continual learning approach using vision transformers. Specifically, we introduce constructive noises at various stages of the vision transformer and enforce consistency in predictions with respect to an exponential moving average of the working model. Our method provides consistent performance gain over raw image and vanilla representation rehearsal on several challenging CL benchmarks, while being memory efficient and robust to natural and adversarial corruptions.

Abstract: We present AvatarReX, a new method for learning NeRF-based full-body avatars from video data. The learnt avatar not only provides expressive control of the body, hands and the face together, but also supports real-time animation and rendering. To this end, we propose a compositional avatar representation, where the body, hands and the face are separately modeled in a way that the structural prior from parametric mesh templates is properly utilized without compromising representation flexibility. Furthermore, we disentangle the geometry and appearance for each part. With these technical designs, we propose a dedicated deferred rendering pipeline, which can be executed in real-time framerate to synthesize high-quality free-view images. The disentanglement of geometry and appearance also allows us to design a two-pass training strategy that combines volume rendering and surface rendering for network training. In this way, patch-level supervision can be applied to force the network to learn sharp appearance details on the basis of geometry estimation. Overall, our method enables automatic construction of expressive full-body avatars with real-time rendering capability, and can generate photo-realistic images with dynamic details for novel body motions and facial expressions.

Abstract: We present a vision and language model named MultiModal-GPT to conduct multi-round dialogue with humans. MultiModal-GPT can follow various instructions from humans, such as generating a detailed caption, counting the number of interested objects, and answering general questions from users. MultiModal-GPT is parameter-efficiently fine-tuned from OpenFlamingo, with Low-rank Adapter (LoRA) added both in the cross-attention part and the self-attention part of the language model. We first construct instruction templates with vision and language data for multi-modality instruction tuning to make the model understand and follow human instructions. We find the quality of training data is vital for the dialogue performance, where few data containing short answers can lead the model to respond shortly to any instructions. To further enhance the ability to chat with humans of the MultiModal-GPT, we utilize language-only instruction-following data to train the MultiModal-GPT jointly. The joint training of language-only and visual-language instructions with the \emph{same} instruction template effectively improves dialogue performance. Various demos show the ability of continuous dialogue of MultiModal-GPT with humans. Code and demo are at https://github.com/open-mmlab/Multimodal-GPT

Abstract: Hand gesture serves as a crucial role during the expression of sign language. Current deep learning based methods for sign language understanding (SLU) are prone to over-fitting due to insufficient sign data resource and suffer limited interpretability. In this paper, we propose the first self-supervised pre-trainable SignBERT+ framework with model-aware hand prior incorporated. In our framework, the hand pose is regarded as a visual token, which is derived from an off-the-shelf detector. Each visual token is embedded with gesture state and spatial-temporal position encoding. To take full advantage of current sign data resource, we first perform self-supervised learning to model its statistics. To this end, we design multi-level masked modeling strategies (joint, frame and clip) to mimic common failure detection cases. Jointly with these masked modeling strategies, we incorporate model-aware hand prior to better capture hierarchical context over the sequence. After the pre-training, we carefully design simple yet effective prediction heads for downstream tasks. To validate the effectiveness of our framework, we perform extensive experiments on three main SLU tasks, involving isolated and continuous sign language recognition (SLR), and sign language translation (SLT). Experimental results demonstrate the effectiveness of our method, achieving new state-of-the-art performance with a notable gain.

Abstract: Assessing the artness of AI-generated images continues to be a challenge within the realm of image generation. Most existing metrics cannot be used to perform instance-level and reference-free artness evaluation. This paper presents ArtScore, a metric designed to evaluate the degree to which an image resembles authentic artworks by artists (or conversely photographs), thereby offering a novel approach to artness assessment. We first blend pre-trained models for photo and artwork generation, resulting in a series of mixed models. Subsequently, we utilize these mixed models to generate images exhibiting varying degrees of artness with pseudo-annotations. Each photorealistic image has a corresponding artistic counterpart and a series of interpolated images that range from realistic to artistic. This dataset is then employed to train a neural network that learns to estimate quantized artness levels of arbitrary images. Extensive experiments reveal that the artness levels predicted by ArtScore align more closely with human artistic evaluation than existing evaluation metrics, such as Gram loss and ArtFID.

Abstract: In order to deal with the sparse and unstructured raw point clouds, LiDAR based 3D object detection research mostly focuses on designing dedicated local point aggregators for fine-grained geometrical modeling. In this paper, we revisit the local point aggregators from the perspective of allocating computational resources. We find that the simplest pillar based models perform surprisingly well considering both accuracy and latency. Additionally, we show that minimal adaptions from the success of 2D object detection, such as enlarging receptive field, significantly boost the performance. Extensive experiments reveal that our pillar based networks with modernized designs in terms of architecture and training render the state-of-the-art performance on the two popular benchmarks: Waymo Open Dataset and nuScenes. Our results challenge the common intuition that the detailed geometry modeling is essential to achieve high performance for 3D object detection.

Abstract: We address the task of estimating 6D camera poses from sparse-view image sets (2-8 images). This task is a vital pre-processing stage for nearly all contemporary (neural) reconstruction algorithms but remains challenging given sparse views, especially for objects with visual symmetries and texture-less surfaces. We build on the recent RelPose framework which learns a network that infers distributions over relative rotations over image pairs. We extend this approach in two key ways; first, we use attentional transformer layers to process multiple images jointly, since additional views of an object may resolve ambiguous symmetries in any given image pair (such as the handle of a mug that becomes visible in a third view). Second, we augment this network to also report camera translations by defining an appropriate coordinate system that decouples the ambiguity in rotation estimation from translation prediction. Our final system results in large improvements in 6D pose prediction over prior art on both seen and unseen object categories and also enables pose estimation and 3D reconstruction for in-the-wild objects.