Computer Vision and Pattern Recognition (cs.CV)

Abstract: Domain generalization person re-identification (DG-ReID) aims to train a model on source domains and generalize well on unseen domains. Vision Transformer usually yields better generalization ability than common CNN networks under distribution shifts. However, Transformer-based ReID models inevitably over-fit to domain-specific biases due to the supervised learning strategy on the source domain. We observe that while the global images of different IDs should have different features, their similar local parts (e.g., black backpack) are not bounded by this constraint. Motivated by this, we propose a pure Transformer model (termed Part-aware Transformer) for DG-ReID by designing a proxy task, named Cross-ID Similarity Learning (CSL), to mine local visual information shared by different IDs. This proxy task allows the model to learn generic features because it only cares about the visual similarity of the parts regardless of the ID labels, thus alleviating the side effect of domain-specific biases. Based on the local similarity obtained in CSL, a Part-guided Self-Distillation (PSD) is proposed to further improve the generalization of global features. Our method achieves state-of-the-art performance under most DG ReID settings. Under the Market$\to$Duke setting, our method exceeds state-of-the-art by 10.9% and 12.8% in Rank1 and mAP, respectively. The code is available at https://github.com/liyuke65535/Part-Aware-Transformer.

Abstract: Image deraining is a challenging task that involves restoring degraded images affected by rain streaks.

Abstract: Near-infrared (NIR) image spectrum translation is a challenging problem with many promising applications. Existing methods struggle with the mapping ambiguity between the NIR and the RGB domains, and generalize poorly due to the limitations of models' learning capabilities and the unavailability of sufficient NIR-RGB image pairs for training. To address these challenges, we propose a cooperative learning paradigm that colorizes NIR images in parallel with another proxy grayscale colorization task by exploring latent cross-domain priors (i.e., latent spectrum context priors and task domain priors), dubbed CoColor. The complementary statistical and semantic spectrum information from these two task domains -- in the forms of pre-trained colorization networks -- are brought in as task domain priors. A bilateral domain translation module is subsequently designed, in which intermittent NIR images are generated from grayscale and colorized in parallel with authentic NIR images; and vice versa for the grayscale images. These intermittent transformations act as latent spectrum context priors for efficient domain knowledge exchange. We progressively fine-tune and fuse these modules with a series of pixel-level and feature-level consistency constraints. Experiments show that our proposed cooperative learning framework produces satisfactory spectrum translation outputs with diverse colors and rich textures, and outperforms state-of-the-art counterparts by 3.95dB and 4.66dB in terms of PNSR for the NIR and grayscale colorization tasks, respectively.

Abstract: This paper proposes a technique for efficiently modeling dynamic humans by explicifying the implicit neural fields via a Neural Explicit Surface (NES). Implicit neural fields have advantages over traditional explicit representations in modeling dynamic 3D content from sparse observations and effectively representing complex geometries and appearances. Implicit neural fields defined in 3D space, however, are expensive to render due to the need for dense sampling during volumetric rendering. Moreover, their memory efficiency can be further optimized when modeling sparse 3D space. To overcome these issues, the paper proposes utilizing Neural Explicit Surface (NES) to explicitly represent implicit neural fields, facilitating memory and computational efficiency. To achieve this, the paper creates a fully differentiable conversion between the implicit neural fields and the explicit rendering interface of NES, leveraging the strengths of both implicit and explicit approaches. This conversion enables effective training of the hybrid representation using implicit methods and efficient rendering by integrating the explicit rendering interface with a newly proposed rasterization-based neural renderer that only incurs a texture color query once for the initial ray interaction with the explicit surface, resulting in improved inference efficiency. NES describes dynamic human geometries with pose-dependent neural implicit surface deformation fields and their dynamic neural textures both in 2D space, which is a more memory-efficient alternative to traditional 3D methods, reducing redundancy and computational load. The comprehensive experiments show that NES performs similarly to previous 3D approaches, with greatly improved rendering speed and reduced memory cost.

Abstract: With the emergence of VR and AR, 360{\deg} data attracts increasing attention from the computer vision and multimedia communities. Typically, 360{\deg} data is projected into 2D ERP (equirectangular projection) images for feature extraction. However, existing methods cannot handle the distortions that result from the projection, hindering the development of 360-data-based tasks. Therefore, in this paper, we propose a Transformer-based model called DATFormer to address the distortion problem. We tackle this issue from two perspectives. Firstly, we introduce two distortion-adaptive modules. The first is a Distortion Mapping Module, which guides the model to pre-adapt to distorted features globally. The second module is a Distortion-Adaptive Attention Block that reduces local distortions on multi-scale features. Secondly, to exploit the unique characteristics of 360{\deg} data, we present a learnable relation matrix and use it as part of the positional embedding to further improve performance. Extensive experiments are conducted on three public datasets, and the results show that our model outperforms existing 2D SOD (salient object detection) and 360 SOD methods.

Abstract: Transformer has recently gained considerable popularity in low-level vision tasks, including image super-resolution (SR). These networks utilize self-attention along different dimensions, spatial or channel, and achieve impressive performance. This inspires us to combine the two dimensions in Transformer for a more powerful representation capability. Based on the above idea, we propose a novel Transformer model, Dual Aggregation Transformer (DAT), for image SR. Our DAT aggregates features across spatial and channel dimensions, in the inter-block and intra-block dual manner. Specifically, we alternately apply spatial and channel self-attention in consecutive Transformer blocks. The alternate strategy enables DAT to capture the global context and realize inter-block feature aggregation. Furthermore, we propose the adaptive interaction module (AIM) and the spatial-gate feed-forward network (SGFN) to achieve intra-block feature aggregation. AIM complements two self-attention mechanisms from corresponding dimensions. Meanwhile, SGFN introduces additional non-linear spatial information in the feed-forward network. Extensive experiments show that our DAT surpasses current methods. Code and models are obtainable at https://github.com/zhengchen1999/DAT.

Abstract: Class-incremental learning (CIL) has achieved remarkable successes in learning new classes consecutively while overcoming catastrophic forgetting on old categories. However, most existing CIL methods unreasonably assume that all old categories have the same forgetting pace, and neglect negative influence of forgetting heterogeneity among different old classes on forgetting compensation. To surmount the above challenges, we develop a novel Heterogeneous Forgetting Compensation (HFC) model, which can resolve heterogeneous forgetting of easy-to-forget and hard-to-forget old categories from both representation and gradient aspects. Specifically, we design a task-semantic aggregation block to alleviate heterogeneous forgetting from representation aspect. It aggregates local category information within each task to learn task-shared global representations. Moreover, we develop two novel plug-and-play losses: a gradient-balanced forgetting compensation loss and a gradient-balanced relation distillation loss to alleviate forgetting from gradient aspect. They consider gradient-balanced compensation to rectify forgetting heterogeneity of old categories and heterogeneous relation consistency. Experiments on several representative datasets illustrate effectiveness of our HFC model. The code is available at https://github.com/JiahuaDong/HFC.

Abstract: Training helps to maintain and improve sufficient muscle function, body control, and body coordination. These are important to reduce the risk of fracture incidents caused by falls, especially for the elderly or people recovering from injury. Virtual reality training can offer a cost-effective and individualized training experience. We present an application for the HoloLens 2 to enable musculoskeletal training for elderly and impaired persons to allow for autonomous training and automatic progress evaluation. We designed a virtual downhill skiing scenario that is controlled by body movement to stimulate balance and body control. By adapting the parameters of the ski slope, we can tailor the intensity of the training to individual users. In this work, we evaluate whether the movement data of the HoloLens 2 alone is sufficient to control and predict body movement and joint angles during musculoskeletal training. We record the movements of 10 healthy volunteers with external tracking cameras and track a set of body and joint angles of the participant during training. We estimate correlation coefficients and systematically analyze whether whole body movement can be derived from the movement data of the HoloLens 2. No participant reports movement sickness effects and all were able to quickly interact and control their movement during skiing. Our results show a high correlation between HoloLens 2 movement data and the external tracking of the upper body movement and joint angles of the lower limbs.

Abstract: Recently, there has been a growing interest in constructing deep learning schemes for Low-Light Vision (LLV). Existing techniques primarily focus on designing task-specific and data-dependent vision models on the standard RGB domain, which inherently contain latent data associations. In this study, we propose a generic low-light vision solution by introducing a generative block to convert data from the RAW to the RGB domain. This novel approach connects diverse vision problems by explicitly depicting data generation, which is the first in the field. To precisely characterize the latent correspondence between the generative procedure and the vision task, we establish a bilevel model with the parameters of the generative block defined as the upper level and the parameters of the vision task defined as the lower level. We further develop two types of learning strategies targeting different goals, namely low cost and high accuracy, to acquire a new bilevel generative learning paradigm. The generative blocks embrace a strong generalization ability in other low-light vision tasks through the bilevel optimization on enhancement tasks. Extensive experimental evaluations on three representative low-light vision tasks, namely enhancement, detection, and segmentation, fully demonstrate the superiority of our proposed approach. The code will be available at https://github.com/Yingchi1998/BGL.

Abstract: The explosive growth of computation and energy cost of artificial intelligence has spurred strong interests in new computing modalities as potential alternatives to conventional electronic processors. Photonic processors that execute operations using photons instead of electrons, have promised to enable optical neural networks with ultra-low latency and power consumption. However, existing optical neural networks, limited by the underlying network designs, have achieved image recognition accuracy much lower than state-of-the-art electronic neural networks. In this work, we close this gap by introducing a large-kernel spatially-varying convolutional neural network learned via low-dimensional reparameterization techniques. We experimentally instantiate the network with a flat meta-optical system that encompasses an array of nanophotonic structures designed to induce angle-dependent responses. Combined with an extremely lightweight electronic backend with approximately 2K parameters we demonstrate a nanophotonic neural network reaches 73.80\% blind test classification accuracy on CIFAR-10 dataset, and, as such, the first time, an optical neural network outperforms the first modern digital neural network -- AlexNet (72.64\%) with 57M parameters, bringing optical neural network into modern deep learning era.

Abstract: Recently, the tokens of images share the same static data flow in many dense networks. However, challenges arise from the variance among the objects in images, such as large variations in the spatial scale and difficulties of recognition for visual entities. In this paper, we propose a data-dependent token routing strategy to elaborate the routing paths of image tokens for Dynamic Vision Transformer, dubbed DiT. The proposed framework generates a data-dependent path per token, adapting to the object scales and visual discrimination of tokens. In feed-forward, the differentiable routing gates are designed to select the scaling paths and feature transformation paths for image tokens, leading to multi-path feature propagation. In this way, the impact of object scales and visual discrimination of image representation can be carefully tuned. Moreover, the computational cost can be further reduced by giving budget constraints to the routing gate and early-stopping of feature extraction. In experiments, our DiT achieves superior performance and favorable complexity/accuracy trade-offs than many SoTA methods on ImageNet classification, object detection, instance segmentation, and semantic segmentation. Particularly, the DiT-B5 obtains 84.8\% top-1 Acc on ImageNet with 10.3 GFLOPs, which is 1.0\% higher than that of the SoTA method with similar computational complexity. These extensive results demonstrate that DiT can serve as versatile backbones for various vision tasks.

Abstract: Obtaining labelled data to train deep learning methods for estimating animal pose is challenging. Recently, synthetic data has been widely used for pose estimation tasks, but most methods still rely on supervised learning paradigms utilising synthetic images and labels. Can training be fully unsupervised? Is a tiny synthetic dataset sufficient? What are the minimum assumptions that we could make for estimating animal pose? Our proposal addresses these questions through a simple yet effective self-supervised method that only assumes the availability of unlabelled images and a small set of synthetic 2D poses. We completely remove the need for any 3D or 2D pose annotations (or complex 3D animal models), and surprisingly our approach can still learn accurate 3D and 2D poses simultaneously. We train our method with unlabelled images of horses mainly collected for YouTube videos and a prior consisting of 2D synthetic poses. The latter is three times smaller than the number of images needed for training. We test our method on a challenging set of horse images and evaluate the predicted 3D and 2D poses. We demonstrate that it is possible to learn accurate animal poses even with as few assumptions as unlabelled images and a small set of 2D poses generated from synthetic data. Given the minimum requirements and the abundance of unlabelled data, our method could be easily deployed to different animals.

Abstract: While current face animation methods can manipulate expressions individually, they suffer from several limitations. The expressions manipulated by some motion-based facial reenactment models are crude. Other ideas modeled with facial action units cannot generalize to arbitrary expressions not covered by annotations. In this paper, we introduce a novel Geometry-aware Facial Expression Translation (GaFET) framework, which is based on parametric 3D facial representations and can stably decoupled expression. Among them, a Multi-level Feature Aligned Transformer is proposed to complement non-geometric facial detail features while addressing the alignment challenge of spatial features. Further, we design a De-expression model based on StyleGAN, in order to reduce the learning difficulty of GaFET in unpaired "in-the-wild" images. Extensive qualitative and quantitative experiments demonstrate that we achieve higher-quality and more accurate facial expression transfer results compared to state-of-the-art methods, and demonstrate applicability of various poses and complex textures. Besides, videos or annotated training data are omitted, making our method easier to use and generalize.

Abstract: Calibration-based methods have dominated RAW image denoising under extremely low-light environments. However, these methods suffer from several main deficiencies: 1) the calibration procedure is laborious and time-consuming, 2) denoisers for different cameras are difficult to transfer, and 3) the discrepancy between synthetic noise and real noise is enlarged by high digital gain. To overcome the above shortcomings, we propose a calibration-free pipeline for Lighting Every Drakness (LED), regardless of the digital gain or camera sensor. Instead of calibrating the noise parameters and training repeatedly, our method could adapt to a target camera only with few-shot paired data and fine-tuning. In addition, well-designed structural modification during both stages alleviates the domain gap between synthetic and real noise without any extra computational cost. With 2 pairs for each additional digital gain (in total 6 pairs) and 0.5% iterations, our method achieves superior performance over other calibration-based methods. Our code is available at https://github.com/Srameo/LED .

Abstract: In recent years, diffusion models have emerged as the most powerful approach in image synthesis. However, applying these models directly to video synthesis presents challenges, as it often leads to noticeable flickering contents. Although recently proposed zero-shot methods can alleviate flicker to some extent, we still struggle to generate coherent videos. In this paper, we propose DiffSynth, a novel approach that aims to convert image synthesis pipelines to video synthesis pipelines. DiffSynth consists of two key components: a latent in-iteration deflickering framework and a video deflickering algorithm. The latent in-iteration deflickering framework applies video deflickering to the latent space of diffusion models, effectively preventing flicker accumulation in intermediate steps. Additionally, we propose a video deflickering algorithm, named patch blending algorithm, that remaps objects in different frames and blends them together to enhance video consistency. One of the notable advantages of DiffSynth is its general applicability to various video synthesis tasks, including text-guided video stylization, fashion video synthesis, image-guided video stylization, video restoring, and 3D rendering. In the task of text-guided video stylization, we make it possible to synthesize high-quality videos without cherry-picking. The experimental results demonstrate the effectiveness of DiffSynth. All videos can be viewed on our project page. Source codes will also be released.

Abstract: This research paper presents a novel approach to pothole detection using Deep Learning and Image Processing techniques. The proposed system leverages the VGG16 model for feature extraction and utilizes a custom Siamese network with triplet loss, referred to as RoadScan. The system aims to address the critical issue of potholes on roads, which pose significant risks to road users. Accidents due to potholes on the roads have led to numerous accidents. Although it is necessary to completely remove potholes, it is a time-consuming process. Hence, a general road user should be able to detect potholes from a safe distance in order to avoid damage. Existing methods for pothole detection heavily rely on object detection algorithms which tend to have a high chance of failure owing to the similarity in structures and textures of a road and a pothole. Additionally, these systems utilize millions of parameters thereby making the model difficult to use in small-scale applications for the general citizen. By analyzing diverse image processing methods and various high-performing networks, the proposed model achieves remarkable performance in accurately detecting potholes. Evaluation metrics such as accuracy, EER, precision, recall, and AUROC validate the effectiveness of the system. Additionally, the proposed model demonstrates computational efficiency and cost-effectiveness by utilizing fewer parameters and data for training. The research highlights the importance of technology in the transportation sector and its potential to enhance road safety and convenience. The network proposed in this model performs with a 96.12 % accuracy, 3.89 % EER, and a 0.988 AUROC value, which is highly competitive with other state-of-the-art works.

Abstract: This paper present a comprehensive comparative analysis of supervised and self-supervised models for deepfake detection. We evaluate eight supervised deep learning architectures and two transformer-based models pre-trained using self-supervised strategies (DINO, CLIP) on four benchmarks (FakeAVCeleb, CelebDF-V2, DFDC, and FaceForensics++). Our analysis includes intra-dataset and inter-dataset evaluations, examining the best performing models, generalisation capabilities, and impact of augmentations. We also investigate the trade-off between model size and performance. Our main goal is to provide insights into the effectiveness of different deep learning architectures (transformers, CNNs), training strategies (supervised, self-supervised), and deepfake detection benchmarks. These insights can help guide the development of more accurate and reliable deepfake detection systems, which are crucial in mitigating the harmful impact of deepfakes on individuals and society.

Abstract: Adversarial attacks can compromise the robustness of real-world detection models. However, evaluating these models under real-world conditions poses challenges due to resource-intensive experiments. Virtual simulations offer an alternative, but the absence of standardized benchmarks hampers progress. Addressing this, we propose an innovative instant-level data generation pipeline using the CARLA simulator. Through this pipeline, we establish the Discrete and Continuous Instant-level (DCI) dataset, enabling comprehensive experiments involving three detection models and three physical adversarial attacks. Our findings highlight diverse model performances under adversarial conditions. Yolo v6 demonstrates remarkable resilience, experiencing just a marginal 6.59% average drop in average precision (AP). In contrast, the ASA attack yields a substantial 14.51% average AP reduction, twice the effect of other algorithms. We also note that static scenes yield higher recognition AP values, and outcomes remain relatively consistent across varying weather conditions. Intriguingly, our study suggests that advancements in adversarial attack algorithms may be approaching its ``limitation''.In summary, our work underscores the significance of adversarial attacks in real-world contexts and introduces the DCI dataset as a versatile benchmark. Our findings provide valuable insights for enhancing the robustness of detection models and offer guidance for future research endeavors in the realm of adversarial attacks.

Abstract: In recent years, weakly supervised models have aided in mass detection using mammography images, decreasing the need for pixel-level annotations. However, most existing models in the literature rely on Class Activation Maps (CAM) as the activation method, overlooking the potential benefits of exploring other activation techniques. This work presents a study that explores and compares different activation maps in conjunction with state-of-the-art methods for weakly supervised training in mammography images. Specifically, we investigate CAM, GradCAM, GradCAM++, XGradCAM, and LayerCAM methods within the framework of the GMIC model for mass detection in mammography images. The evaluation is conducted on the VinDr-Mammo dataset, utilizing the metrics Accuracy, True Positive Rate (TPR), False Negative Rate (FNR), and False Positive Per Image (FPPI). Results show that using different strategies of activation maps during training and test stages leads to an improvement of the model. With this strategy, we improve the results of the GMIC method, decreasing the FPPI value and increasing TPR.

Abstract: We propose a novel framework to automatically learn to aggregate and transform photometric measurements from multiple unstructured views into spatially distinctive and view-invariant low-level features, which are fed to a multi-view stereo method to enhance 3D reconstruction. The illumination conditions during acquisition and the feature transform are jointly trained on a large amount of synthetic data. We further build a system to reconstruct the geometry and anisotropic reflectance of a variety of challenging objects from hand-held scans. The effectiveness of the system is demonstrated with a lightweight prototype, consisting of a camera and an array of LEDs, as well as an off-the-shelf tablet. Our results are validated against reconstructions from a professional 3D scanner and photographs, and compare favorably with state-of-the-art techniques.

Abstract: For a machine learning model to generalize effectively to unseen data within a particular problem domain, it is well-understood that the data needs to be of sufficient size and representative of real-world scenarios. Nonetheless, real-world datasets frequently have overrepresented and underrepresented groups. One solution to mitigate bias in machine learning is to leverage a diverse and representative dataset. Training a model on a dataset that covers all demographics is crucial to reducing bias in machine learning. However, collecting and labeling large-scale datasets has been challenging, prompting the use of synthetic data generation and active labeling to decrease the costs of manual labeling. The focus of this study was to generate a robust face image dataset using the StyleGAN model. In order to achieve a balanced distribution of the dataset among different demographic groups, a synthetic dataset was created by controlling the generation process of StyleGaN and annotated for different downstream tasks.

Abstract: Recent advances in segmentation-free keyword spotting treat this problem w.r.t. an object detection paradigm and borrow from state-of-the-art detection systems to simultaneously propose a word bounding box proposal mechanism and compute a corresponding representation. Contrary to the norm of such methods that rely on complex and large DNN models, we propose a novel segmentation-free system that efficiently scans a document image to find rectangular areas that include the query information. The underlying model is simple and compact, predicting character occurrences over rectangular areas through an implicitly learned scale map, trained on word-level annotated images. The proposed document scanning is then performed using this character counting in a cost-effective manner via integral images and binary search. Finally, the retrieval similarity by character counting is refined by a pyramidal representation and a CTC-based re-scoring algorithm, fully utilizing the trained CNN model. Experimental validation on two widely-used datasets shows that our method achieves state-of-the-art results outperforming the more complex alternatives, despite the simplicity of the underlying model.

Abstract: Recent interactive segmentation methods iteratively take source image, user guidance and previously predicted mask as the input without considering the invariant nature of the source image. As a result, extracting features from the source image is repeated in each interaction, resulting in substantial computational redundancy. In this work, we propose the Feature Decoupling-Recycling Network (FDRN), which decouples the modeling components based on their intrinsic discrepancies and then recycles components for each user interaction. Thus, the efficiency of the whole interactive process can be significantly improved. To be specific, we apply the Decoupling-Recycling strategy from three perspectives to address three types of discrepancies, respectively. First, our model decouples the learning of source image semantics from the encoding of user guidance to process two types of input domains separately. Second, FDRN decouples high-level and low-level features from stratified semantic representations to enhance feature learning. Third, during the encoding of user guidance, current user guidance is decoupled from historical guidance to highlight the effect of current user guidance. We conduct extensive experiments on 6 datasets from different domains and modalities, which demonstrate the following merits of our model: 1) superior efficiency than other methods, particularly advantageous in challenging scenarios requiring long-term interactions (up to 4.25x faster), while achieving favorable segmentation performance; 2) strong applicability to various methods serving as a universal enhancement technique; 3) well cross-task generalizability, e.g., to medical image segmentation, and robustness against misleading user guidance.

Abstract: Supervised deep learning models require significant amount of labelled data to achieve an acceptable performance on a specific task. However, when tested on unseen data, the models may not perform well. Therefore, the models need to be trained with additional and varying labelled data to improve the generalization. In this work, our goal is to understand the models, their performance and generalization. We establish image-image, dataset-dataset, and image-dataset distances to gain insights into the model's behavior. Our proposed distance metric when combined with model performance can help in selecting an appropriate model/architecture from a pool of candidate architectures. We have shown that the generalization of these models can be improved by only adding a small number of unseen images (say 1, 3 or 7) into the training set. Our proposed approach reduces training and annotation costs while providing an estimate of model performance on unseen data in dynamic environments.

Abstract: Digital soil mapping (DSM) is an advanced approach that integrates statistical modeling and cutting-edge technologies, including machine learning (ML) methods, to accurately depict soil properties and their spatial distribution. Soil organic carbon (SOC) is a crucial soil attribute providing valuable insights into soil health, nutrient cycling, greenhouse gas emissions, and overall ecosystem productivity. This study highlights the significance of spatial-temporal deep learning (DL) techniques within the DSM framework. A novel architecture is proposed, incorporating spatial information using a base convolutional neural network (CNN) model and spatial attention mechanism, along with climate temporal information using a long short-term memory (LSTM) network, for SOC prediction across Europe. The model utilizes a comprehensive set of environmental features, including Landsat-8 images, topography, remote sensing indices, and climate time series, as input features. Results demonstrate that the proposed framework outperforms conventional ML approaches like random forest commonly used in DSM, yielding lower root mean square error (RMSE). This model is a robust tool for predicting SOC and could be applied to other soil properties, thereby contributing to the advancement of DSM techniques and facilitating land management and decision-making processes based on accurate information.

Abstract: Extracting useful visual cues for the downstream tasks is especially challenging under low-light vision. Prior works create enhanced representations by either correlating visual quality with machine perception or designing illumination-degrading transformation methods that require pre-training on synthetic datasets. We argue that optimizing enhanced image representation pertaining to the loss of the downstream task can result in more expressive representations. Therefore, in this work, we propose a novel module, FeatEnHancer, that hierarchically combines multiscale features using multiheaded attention guided by task-related loss function to create suitable representations. Furthermore, our intra-scale enhancement improves the quality of features extracted at each scale or level, as well as combines features from different scales in a way that reflects their relative importance for the task at hand. FeatEnHancer is a general-purpose plug-and-play module and can be incorporated into any low-light vision pipeline. We show with extensive experimentation that the enhanced representation produced with FeatEnHancer significantly and consistently improves results in several low-light vision tasks, including dark object detection (+5.7 mAP on ExDark), face detection (+1.5 mAPon DARK FACE), nighttime semantic segmentation (+5.1 mIoU on ACDC ), and video object detection (+1.8 mAP on DarkVision), highlighting the effectiveness of enhancing hierarchical features under low-light vision.

Abstract: Self-supervised video denoising has seen decent progress through the use of blind spot networks. However, under their blind spot constraints, previous self-supervised video denoising methods suffer from significant information loss and texture destruction in either the whole reference frame or neighbor frames, due to their inadequate consideration of the receptive field. Moreover, the limited number of available neighbor frames in previous methods leads to the discarding of distant temporal information. Nonetheless, simply adopting existing recurrent frameworks does not work, since they easily break the constraints on the receptive field imposed by self-supervision. In this paper, we propose RDRF for self-supervised video denoising, which not only fully exploits both the reference and neighbor frames with a denser receptive field, but also better leverages the temporal information from both local and distant neighbor features. First, towards a comprehensive utilization of information from both reference and neighbor frames, RDRF realizes a denser receptive field by taking more neighbor pixels along the spatial and temporal dimensions. Second, it features a self-supervised recurrent video denoising framework, which concurrently integrates distant and near-neighbor temporal features. This enables long-term bidirectional information aggregation, while mitigating error accumulation in the plain recurrent framework. Our method exhibits superior performance on both synthetic and real video denoising datasets. Codes will be available at https://github.com/Wang-XIaoDingdd/RDRF.

Abstract: Creating expressive, diverse and high-quality 3D avatars from highly customized text descriptions and pose guidance is a challenging task, due to the intricacy of modeling and texturing in 3D that ensure details and various styles (realistic, fictional, etc). We present AvatarVerse, a stable pipeline for generating expressive high-quality 3D avatars from nothing but text descriptions and pose guidance. In specific, we introduce a 2D diffusion model conditioned on DensePose signal to establish 3D pose control of avatars through 2D images, which enhances view consistency from partially observed scenarios. It addresses the infamous Janus Problem and significantly stablizes the generation process. Moreover, we propose a progressive high-resolution 3D synthesis strategy, which obtains substantial improvement over the quality of the created 3D avatars. To this end, the proposed AvatarVerse pipeline achieves zero-shot 3D modeling of 3D avatars that are not only more expressive, but also in higher quality and fidelity than previous works. Rigorous qualitative evaluations and user studies showcase AvatarVerse's superiority in synthesizing high-fidelity 3D avatars, leading to a new standard in high-quality and stable 3D avatar creation. Our project page is: https://avatarverse3d.github.io

Abstract: Retrofitting and thermographic survey (TS) companies in Scotland collaborate with social housing providers to tackle fuel poverty. They employ ground-level infrared (IR) camera-based-TSs (GIRTSs) for collecting thermal images to identi-fy the heat loss sources resulting from poor insulation. However, this identifica-tion process is labor-intensive and time-consuming, necessitating extensive data processing. To automate this, an AI-driven approach is necessary. Therefore, this study proposes a deep learning (DL)-based segmentation framework using the Mask Region Proposal Convolutional Neural Network (Mask RCNN) to validate its applicability to these thermal images. The objective of the framework is to au-tomatically identify, and crop heat loss sources caused by weak insulation, while also eliminating obstructive objects present in those images. By doing so, it min-imizes labor-intensive tasks and provides an automated, consistent, and reliable solution. To validate the proposed framework, approximately 2500 thermal imag-es were collected in collaboration with industrial TS partner. Then, 1800 repre-sentative images were carefully selected with the assistance of experts and anno-tated to highlight the target objects (TO) to form the final dataset. Subsequently, a transfer learning strategy was employed to train the dataset, progressively aug-menting the training data volume and fine-tuning the pre-trained baseline Mask RCNN. As a result, the final fine-tuned model achieved a mean average precision (mAP) score of 77.2% for segmenting the TO, demonstrating the significant po-tential of proposed framework in accurately quantifying energy loss in Scottish homes.

Abstract: Accurate catheter tracking is crucial during minimally invasive endovascular procedures (MIEP), and electromagnetic (EM) tracking is a widely used technology that serves this purpose. However, registration between preoperative images and the EM tracking system is often challenging. Existing registration methods typically require manual interactions, which can be time-consuming, increase the risk of errors and change the procedural workflow. Although several registration methods are available for catheter tracking, such as marker-based and path-based approaches, their limitations can impact the accuracy of the resulting tracking solution, consequently, the outcome of the medical procedure. This paper introduces a novel automated catheter registration method for EM-guided MIEP. The method utilizes 3D signal temporal analysis, such as Dynamic Time Warping (DTW) algorithms, to improve registration accuracy and reliability compared to existing methods. DTW can accurately warp and match EM-tracked paths to the vessel's centerline, making it particularly suitable for registration. The introduced registration method is evaluated for accuracy in a vascular phantom using a marker-based registration as the ground truth. The results indicate that the DTW method yields accurate and reliable registration outcomes, with a mean error of $2.22$mm. The introduced registration method presents several advantages over state-of-the-art methods, such as high registration accuracy, no initialization required, and increased automation.

Abstract: Cryo-electron microscopy (cryo-EM) is a technique for reconstructing the 3-dimensional (3D) structure of biomolecules (especially large protein complexes and molecular assemblies). As the resolution increases to the near-atomic scale, building protein structures de novo from cryo-EM maps becomes possible. Recently, recognition-based de novo building methods have shown the potential to streamline this process. However, it cannot build a complete structure due to the low signal-to-noise ratio (SNR) problem. At the same time, AlphaFold has led to a great breakthrough in predicting protein structures. This has inspired us to combine fragment recognition and structure prediction methods to build a complete structure. In this paper, we propose a new method named FFF that bridges protein structure prediction and protein structure recognition with flexible fitting. First, a multi-level recognition network is used to capture various structural features from the input 3D cryo-EM map. Next, protein structural fragments are generated using pseudo peptide vectors and a protein sequence alignment method based on these extracted features. Finally, a complete structural model is constructed using the predicted protein fragments via flexible fitting. Based on our benchmark tests, FFF outperforms the baseline methods for building complete protein structures.

Abstract: Real world uses of deep learning require predictable model behavior under distribution shifts. Models such as CLIP show emergent natural distributional robustness comparable to humans, but may require hundreds of millions of training samples. Can we train robust learners in a domain where data is limited? To rigorously address this question, we introduce JANuS (Joint Annotations and Names Set), a collection of four new training datasets with images, labels, and corresponding captions, and perform a series of carefully controlled investigations of factors contributing to robustness in image classification, then compare those results to findings derived from a large-scale meta-analysis. Using this approach, we show that standard ResNet-50 trained with the cross-entropy loss on 2.4 million image samples can attain comparable robustness to a CLIP ResNet-50 trained on 400 million samples. To our knowledge, this is the first result showing (near) state-of-the-art distributional robustness on limited data budgets. Our dataset is available at \url{https://huggingface.co/datasets/penfever/JANuS_dataset}, and the code used to reproduce our experiments can be found at \url{https://github.com/penfever/vlhub/}.

Abstract: Fusarium head blight is a devastating disease that causes significant economic losses annually on small grains. Efficiency, accuracy, and timely detection of FHB in the resistance screening are critical for wheat and barley breeding programs. In recent years, various image processing techniques have been developed using supervised machine learning algorithms for the early detection of FHB. The state-of-the-art convolutional neural network-based methods, such as U-Net, employ a series of encoding blocks to create a local representation and a series of decoding blocks to capture the semantic relations. However, these methods are not often capable of long-range modeling dependencies inside the input data, and their ability to model multi-scale objects with significant variations in texture and shape is limited. Vision transformers as alternative architectures with innate global self-attention mechanisms for sequence-to-sequence prediction, due to insufficient low-level details, may also limit localization capabilities. To overcome these limitations, a new Context Bridge is proposed to integrate the local representation capability of the U-Net network in the transformer model. In addition, the standard attention mechanism of the original transformer is replaced with Efficient Self-attention, which is less complicated than other state-of-the-art methods. To train the proposed network, 12,000 wheat images from an FHB-inoculated wheat field at the SDSU research farm in Volga, SD, were captured. In addition to healthy and unhealthy plants, these images encompass various stages of the disease. A team of expert pathologists annotated the images for training and evaluating the developed model. As a result, the effectiveness of the transformer-based method for FHB-disease detection, through extensive experiments across typical tasks for plant image segmentation, is demonstrated.

Abstract: Recent advances in foundation models present new opportunities for interpretable visual recognition -- one can first query Large Language Models (LLMs) to obtain a set of attributes that describe each class, then apply vision-language models to classify images via these attributes. Pioneering work shows that querying thousands of attributes can achieve performance competitive with image features. However, our further investigation on 8 datasets reveals that LLM-generated attributes in a large quantity perform almost the same as random words. This surprising finding suggests that significant noise may be present in these attributes. We hypothesize that there exist subsets of attributes that can maintain the classification performance with much smaller sizes, and propose a novel learning-to-search method to discover those concise sets of attributes. As a result, on the CUB dataset, our method achieves performance close to that of massive LLM-generated attributes (e.g., 10k attributes for CUB), yet using only 32 attributes in total to distinguish 200 bird species. Furthermore, our new paradigm demonstrates several additional benefits: higher interpretability and interactivity for humans, and the ability to summarize knowledge for a recognition task.

Abstract: Video-based person Re-Identification (V-ReID) aims to retrieve specific persons from raw videos captured by non-overlapped cameras. As a fundamental task, it spreads many multimedia and computer vision applications. However, due to the variations of persons and scenes, there are still many obstacles that must be overcome for high performance. In this work, we notice that both the long-term and short-term information of persons are important for robust video representations. Thus, we propose a novel deep learning framework named Long Short-Term Representation Learning (LSTRL) for effective V-ReID. More specifically, to extract long-term representations, we propose a Multi-granularity Appearance Extractor (MAE), in which four granularity appearances are effectively captured across multiple frames. Meanwhile, to extract short-term representations, we propose a Bi-direction Motion Estimator (BME), in which reciprocal motion information is efficiently extracted from consecutive frames. The MAE and BME are plug-and-play and can be easily inserted into existing networks for efficient feature learning. As a result, they significantly improve the feature representation ability for V-ReID. Extensive experiments on three widely used benchmarks show that our proposed approach can deliver better performances than most state-of-the-arts.

Abstract: Existing polyp segmentation models from colonoscopy images often fail to provide reliable segmentation results on datasets from different centers, limiting their applicability. Our objective in this study is to create a robust and well-generalized segmentation model named PrototypeLab that can assist in polyp segmentation. To achieve this, we incorporate various lighting modes such as White light imaging (WLI), Blue light imaging (BLI), Linked color imaging (LCI), and Flexible spectral imaging color enhancement (FICE) into our new segmentation model, that learns to create prototypes for each class of object present in the images. These prototypes represent the characteristic features of the objects, such as their shape, texture, color. Our model is designed to perform effectively on out-of-distribution (OOD) datasets from multiple centers. We first generate a coarse mask that is used to learn prototypes for the main object class, which are then employed to generate the final segmentation mask. By using prototypes to represent the main class, our approach handles the variability present in the medical images and generalize well to new data since prototype capture the underlying distribution of the data. PrototypeLab offers a promising solution with a dice coefficient of $\geq$ 90\% and mIoU $\geq$ 85\% with a near real-time processing speed for polyp segmentation. It achieved superior performance on OOD datasets compared to 16 state-of-the-art image segmentation architectures, potentially improving clinical outcomes. Codes are available at https://github.com/xxxxx/PrototypeLab.

Abstract: This paper asks whether current self-supervised learning methods, if sufficiently scaled up, would be able to reach human-level visual object recognition capabilities with the same type and amount of visual experience humans learn from. Previous work on this question only considered the scaling of data size. Here, we consider the simultaneous scaling of data size, model size, and image resolution. We perform a scaling experiment with vision transformers up to 633M parameters in size (ViT-H/14) trained with up to 5K hours of human-like video data (long, continuous, mostly egocentric videos) with image resolutions of up to 476x476 pixels. The efficiency of masked autoencoders (MAEs) as a self-supervised learning algorithm makes it possible to run this scaling experiment on an unassuming academic budget. We find that it is feasible to reach human-level object recognition capacity at sub-human scales of model size, data size, and image size, if these factors are scaled up simultaneously. To give a concrete example, we estimate that a 2.5B parameter ViT model trained with 20K hours (2.3 years) of human-like video data with a spatial resolution of 952x952 pixels should be able to reach roughly human-level accuracy on ImageNet. Human-level competence is thus achievable for a fundamental perceptual capability from human-like perceptual experience (human-like in both amount and type) with extremely generic learning algorithms and architectures and without any substantive inductive biases.

Abstract: Peripheral nerve blocks are crucial to treatment of post-surgical pain and are associated with reduction in perioperative opioid use and hospital stay. Accurate interpretation of sono-anatomy is critical for the success of ultrasound (US) guided peripheral nerve blocks and can be challenging to the new operators. This prospective study enrolled 227 subjects who were systematically scanned for supraclavicular and interscalene brachial plexus in various settings using three different US machines to create a dataset of 227 unique videos. In total, 41,000 video frames were annotated by experienced anaesthesiologists using partial automation with object tracking and active contour algorithms. Four baseline neural network models were trained on the dataset and their performance was evaluated for object detection and segmentation tasks. Generalizability of the best suited model was then tested on the datasets constructed from separate US scanners with and without fine-tuning. The results demonstrate that deep learning models can be leveraged for real time segmentation of supraclavicular brachial plexus in neck ultrasonography videos with high accuracy and reliability. Model was also tested for its ability to differentiate between supraclavicular and adjoining interscalene brachial plexus. The entire dataset has been released publicly for further study by the research community.

Abstract: Temporal Sentence Grounding in Videos (TSGV) aims to detect the event timestamps described by the natural language query from untrimmed videos. This paper discusses the challenge of achieving efficient computation in TSGV models while maintaining high performance. Most existing approaches exquisitely design complex architectures to improve accuracy with extra layers and loss, suffering from inefficiency and heaviness. Although some works have noticed that, they only make an issue of feature fusion layers, which can hardly enjoy the highspeed merit in the whole clunky network. To tackle this problem, we propose a novel efficient multi-teacher model (EMTM) based on knowledge distillation to transfer diverse knowledge from both heterogeneous and isomorphic networks. Specifically, We first unify different outputs of the heterogeneous models into one single form. Next, a Knowledge Aggregation Unit (KAU) is built to acquire high-quality integrated soft labels from multiple teachers. After that, the KAU module leverages the multi-scale video and global query information to adaptively determine the weights of different teachers. A Shared Encoder strategy is then proposed to solve the problem that the student shallow layers hardly benefit from teachers, in which an isomorphic teacher is collaboratively trained with the student to align their hidden states. Extensive experimental results on three popular TSGV benchmarks demonstrate that our method is both effective and efficient without bells and whistles.

Abstract: Segmentation is a fundamental problem in surgical scene analysis using artificial intelligence. However, the inherent data scarcity in this domain makes it challenging to adapt traditional segmentation techniques for this task. To tackle this issue, current research employs pretrained models and finetunes them on the given data. Even so, these require training deep networks with millions of parameters every time new data becomes available. A recently published foundation model, Segment-Anything (SAM), generalizes well to a large variety of natural images, hence tackling this challenge to a reasonable extent. However, SAM does not generalize well to the medical domain as is without utilizing a large amount of compute resources for fine-tuning and using task-specific prompts. Moreover, these prompts are in the form of bounding-boxes or foreground/background points that need to be annotated explicitly for every image, making this solution increasingly tedious with higher data size. In this work, we propose AdaptiveSAM - an adaptive modification of SAM that can adjust to new datasets quickly and efficiently, while enabling text-prompted segmentation. For finetuning AdaptiveSAM, we propose an approach called bias-tuning that requires a significantly smaller number of trainable parameters than SAM (less than 2\%). At the same time, AdaptiveSAM requires negligible expert intervention since it uses free-form text as prompt and can segment the object of interest with just the label name as prompt. Our experiments show that AdaptiveSAM outperforms current state-of-the-art methods on various medical imaging datasets including surgery, ultrasound and X-ray. Code is available at https://github.com/JayParanjape/biastuning

Abstract: Recent advancements in Large Vision-Language Models (LVLMs) have demonstrated significant progress in tackling complex multimodal tasks. Among these cutting-edge developments, Google's Bard stands out for its remarkable multimodal capabilities, promoting comprehensive comprehension and reasoning across various domains. This work presents an early and holistic evaluation of LVLMs' multimodal abilities, with a particular focus on Bard, by proposing a lightweight variant of LVLM-eHub, named Tiny LVLM-eHub. In comparison to the vanilla version, Tiny LVLM-eHub possesses several appealing properties. Firstly, it provides a systematic assessment of six categories of multimodal capabilities, including visual perception, visual knowledge acquisition, visual reasoning, visual commonsense, object hallucination, and embodied intelligence, through quantitative evaluation of $42$ standard text-related visual benchmarks. Secondly, it conducts an in-depth analysis of LVLMs' predictions using the ChatGPT Ensemble Evaluation (CEE), which leads to a robust and accurate evaluation and exhibits improved alignment with human evaluation compared to the word matching approach. Thirdly, it comprises a mere $2.1$K image-text pairs, facilitating ease of use for practitioners to evaluate their own offline LVLMs. Through extensive experimental analysis, this study demonstrates that Bard outperforms previous LVLMs in most multimodal capabilities except object hallucination, to which Bard is still susceptible. Tiny LVLM-eHub serves as a baseline evaluation for various LVLMs and encourages innovative strategies aimed at advancing multimodal techniques. Our project is publicly available at \url{https://github.com/OpenGVLab/Multi-Modality-Arena}.

Abstract: Salient Object Detection (SOD) aims to identify and segment the most conspicuous objects in an image or video. As an important pre-processing step, it has many potential applications in multimedia and vision tasks. With the advance of imaging devices, SOD with high-resolution images is of great demand, recently. However, traditional SOD methods are largely limited to low-resolution images, making them difficult to adapt to the development of High-Resolution SOD (HRSOD). Although some HRSOD methods emerge, there are no large enough datasets for training and evaluating. Besides, current HRSOD methods generally produce incomplete object regions and irregular object boundaries. To address above issues, in this work, we first propose a new HRS10K dataset, which contains 10,500 high-quality annotated images at 2K-8K resolution. As far as we know, it is the largest dataset for the HRSOD task, which will significantly help future works in training and evaluating models. Furthermore, to improve the HRSOD performance, we propose a novel Recurrent Multi-scale Transformer (RMFormer), which recurrently utilizes shared Transformers and multi-scale refinement architectures. Thus, high-resolution saliency maps can be generated with the guidance of lower-resolution predictions. Extensive experiments on both high-resolution and low-resolution benchmarks show the effectiveness and superiority of the proposed framework. The source code and dataset are released at: https://github.com/DrowsyMon/RMFormer.

Abstract: In this paper, we aim to study how to build a strong instance segmenter with minimal training time and GPUs, as opposed to the majority of current approaches that pursue more accurate instance segmenter by building more advanced frameworks at the cost of longer training time and higher GPU requirements. To achieve this, we introduce a simple and general framework, termed Mask Frozen-DETR, which can convert any existing DETR-based object detection model into a powerful instance segmentation model. Our method only requires training an additional lightweight mask network that predicts instance masks within the bounding boxes given by a frozen DETR-based object detector. Remarkably, our method outperforms the state-of-the-art instance segmentation method Mask DINO in terms of performance on the COCO test-dev split (55.3% vs. 54.7%) while being over 10X times faster to train. Furthermore, all of our experiments can be trained using only one Tesla V100 GPU with 16 GB of memory, demonstrating the significant efficiency of our proposed framework.

Abstract: LiDAR-based fully sparse architecture has garnered increasing attention. FSDv1 stands out as a representative work, achieving impressive efficacy and efficiency, albeit with intricate structures and handcrafted designs. In this paper, we present FSDv2, an evolution that aims to simplify the previous FSDv1 while eliminating the inductive bias introduced by its handcrafted instance-level representation, thus promoting better general applicability. To this end, we introduce the concept of \textbf{virtual voxels}, which takes over the clustering-based instance segmentation in FSDv1. Virtual voxels not only address the notorious issue of the Center Feature Missing problem in fully sparse detectors but also endow the framework with a more elegant and streamlined approach. Consequently, we develop a suite of components to complement the virtual voxel concept, including a virtual voxel encoder, a virtual voxel mixer, and a virtual voxel assignment strategy. Through empirical validation, we demonstrate that the virtual voxel mechanism is functionally similar to the handcrafted clustering in FSDv1 while being more general. We conduct experiments on three large-scale datasets: Waymo Open Dataset, Argoverse 2 dataset, and nuScenes dataset. Our results showcase state-of-the-art performance on all three datasets, highlighting the superiority of FSDv2 in long-range scenarios and its general applicability to achieve competitive performance across diverse scenarios. Moreover, we provide comprehensive experimental analysis to elucidate the workings of FSDv2. To foster reproducibility and further research, we have open-sourced FSDv2 at https://github.com/tusen-ai/SST.

Abstract: Motion magnification helps us visualize subtle, imperceptible motion. However, prior methods only work for 2D videos captured with a fixed camera. We present a 3D motion magnification method that can magnify subtle motions from scenes captured by a moving camera, while supporting novel view rendering. We represent the scene with time-varying radiance fields and leverage the Eulerian principle for motion magnification to extract and amplify the variation of the embedding of a fixed point over time. We study and validate our proposed principle for 3D motion magnification using both implicit and tri-plane-based radiance fields as our underlying 3D scene representation. We evaluate the effectiveness of our method on both synthetic and real-world scenes captured under various camera setups.

Abstract: We introduce a high throughput 3D scanning solution specifically designed to precisely measure cattle phenotypes. This scanner leverages an array of depth sensors, i.e. time-of-flight (Tof) sensors, each governed by dedicated embedded devices. The system excels at generating high-fidelity 3D point clouds, thus facilitating an accurate mesh that faithfully reconstructs the cattle geometry on the fly. In order to evaluate the performance of our system, we have implemented a two-fold validation process. Initially, we test the scanner's competency in determining volume and surface area measurements within a controlled environment featuring known objects. Secondly, we explore the impact and necessity of multi-device synchronization when operating a series of time-of-flight sensors. Based on the experimental results, the proposed system is capable of producing high-quality meshes of untamed cattle for livestock studies.

Abstract: Learning-based image deraining methods have made great progress. However, the lack of large-scale high-quality paired training samples is the main bottleneck to hamper the real image deraining (RID). To address this dilemma and advance RID, we construct a Large-scale High-quality Paired real rain benchmark (LHP-Rain), including 3000 video sequences with 1 million high-resolution (1920*1080) frame pairs. The advantages of the proposed dataset over the existing ones are three-fold: rain with higher-diversity and larger-scale, image with higher-resolution and higher-quality ground-truth. Specifically, the real rains in LHP-Rain not only contain the classical rain streak/veiling/occlusion in the sky, but also the \textbf{splashing on the ground} overlooked by deraining community. Moreover, we propose a novel robust low-rank tensor recovery model to generate the GT with better separating the static background from the dynamic rain. In addition, we design a simple transformer-based single image deraining baseline, which simultaneously utilize the self-attention and cross-layer attention within the image and rain layer with discriminative feature representation. Extensive experiments verify the superiority of the proposed dataset and deraining method over state-of-the-art.

Abstract: Developability refers to the process of creating a surface without any tearing or shearing from a two-dimensional plane. It finds practical applications in the fabrication industry. An essential characteristic of a developable 3D surface is its zero Gaussian curvature, which means that either one or both of the principal curvatures are zero. This paper introduces a method for reconstructing an approximate developable surface from a neural implicit surface. The central idea of our method involves incorporating a regularization term that operates on the second-order derivatives of the neural implicits, effectively promoting zero Gaussian curvature. Implicit surfaces offer the advantage of smoother deformation with infinite resolution, overcoming the high polygonal constraints of state-of-the-art methods using discrete representations. We draw inspiration from the properties of surface curvature and employ rank minimization techniques derived from compressed sensing. Experimental results on both developable and non-developable surfaces, including those affected by noise, validate the generalizability of our method.

Abstract: We present a Multimodal Backdoor Defense technique TIJO (Trigger Inversion using Joint Optimization). Recent work arXiv:2112.07668 has demonstrated successful backdoor attacks on multimodal models for the Visual Question Answering task. Their dual-key backdoor trigger is split across two modalities (image and text), such that the backdoor is activated if and only if the trigger is present in both modalities. We propose TIJO that defends against dual-key attacks through a joint optimization that reverse-engineers the trigger in both the image and text modalities. This joint optimization is challenging in multimodal models due to the disconnected nature of the visual pipeline which consists of an offline feature extractor, whose output is then fused with the text using a fusion module. The key insight enabling the joint optimization in TIJO is that the trigger inversion needs to be carried out in the object detection box feature space as opposed to the pixel space. We demonstrate the effectiveness of our method on the TrojVQA benchmark, where TIJO improves upon the state-of-the-art unimodal methods from an AUC of 0.6 to 0.92 on multimodal dual-key backdoors. Furthermore, our method also improves upon the unimodal baselines on unimodal backdoors. We present ablation studies and qualitative results to provide insights into our algorithm such as the critical importance of overlaying the inverted feature triggers on all visual features during trigger inversion. The prototype implementation of TIJO is available at https://github.com/SRI-CSL/TIJO.

Abstract: Video Action Recognition (VAR) is a challenging task due to its inherent complexities. Though different approaches have been explored in the literature, designing a unified framework to recognize a large number of human actions is still a challenging problem. Recently, Multi-Modal Learning (MML) has demonstrated promising results in this domain. In literature, 2D skeleton or pose modality has often been used for this task, either independently or in conjunction with the visual information (RGB modality) present in videos. However, the combination of pose, visual information, and text attributes has not been explored yet, though text and pose attributes independently have been proven to be effective in numerous computer vision tasks. In this paper, we present the first pose augmented Vision-language model (VLM) for VAR. Notably, our scheme achieves an accuracy of 92.81% and 73.02% on two popular human video action recognition benchmark datasets, UCF-101 and HMDB-51, respectively, even without any video data pre-training, and an accuracy of 96.11% and 75.75% after kinetics pre-training.

Abstract: We propose an exhaustive methodology that leverages all levels of feature abstraction, targeting an enhancement in the generalizability of image classification to unobserved hospitals. Our approach incorporates augmentation-based self-supervision with common distribution shifts in histopathology scenarios serving as the pretext task. This enables us to derive invariant features from training images without relying on training labels, thereby covering different abstraction levels. Moving onto the subsequent abstraction level, we employ a domain alignment module to facilitate further extraction of invariant features across varying training hospitals. To represent the highly specific features of participating hospitals, an encoder is trained to classify hospital labels, independent of their diagnostic labels. The features from each of these encoders are subsequently disentangled to minimize redundancy and segregate the features. This representation, which spans a broad spectrum of semantic information, enables the development of a model demonstrating increased robustness to unseen images from disparate distributions. Experimental results from the PACS dataset (a domain generalization benchmark), a synthetic dataset created by applying histopathology-specific jitters to the MHIST dataset (defining different domains with varied distribution shifts), and a Renal Cell Carcinoma dataset derived from four image repositories from TCGA, collectively indicate that our proposed model is adept at managing varying levels of image granularity. Thus, it shows improved generalizability when faced with new, out-of-distribution hospital images.

Abstract: Auto-white balance (AWB) correction is a critical operation in image signal processors for accurate and consistent color correction across various illumination scenarios. This paper presents a novel and efficient AWB correction method that achieves at least 35 times faster processing with equivalent or superior performance on high-resolution images for the current state-of-the-art methods. Inspired by deterministic color style transfer, our approach introduces deterministic illumination color mapping, leveraging learnable projection matrices for both canonical illumination form and AWB-corrected output. It involves feeding high-resolution images and corresponding latent representations into a mapping module to derive a canonical form, followed by another mapping module that maps the pixel values to those for the corrected version. This strategy is designed as resolution-agnostic and also enables seamless integration of any pre-trained AWB network as the backbone. Experimental results confirm the effectiveness of our approach, revealing significant performance improvements and reduced time complexity compared to state-of-the-art methods. Our method provides an efficient deep learning-based AWB correction solution, promising real-time, high-quality color correction for digital imaging applications. Source code is available at https://github.com/birdortyedi/DeNIM/

Abstract: Zero-shot skeleton-based action recognition aims to recognize actions of unseen categories after training on data of seen categories. The key is to build the connection between visual and semantic space from seen to unseen classes. Previous studies have primarily focused on encoding sequences into a singular feature vector, with subsequent mapping the features to an identical anchor point within the embedded space. Their performance is hindered by 1) the ignorance of the global visual/semantic distribution alignment, which results in a limitation to capture the true interdependence between the two spaces. 2) the negligence of temporal information since the frame-wise features with rich action clues are directly pooled into a single feature vector. We propose a new zero-shot skeleton-based action recognition method via mutual information (MI) estimation and maximization. Specifically, 1) we maximize the MI between visual and semantic space for distribution alignment; 2) we leverage the temporal information for estimating the MI by encouraging MI to increase as more frames are observed. Extensive experiments on three large-scale skeleton action datasets confirm the effectiveness of our method. Code: https://github.com/YujieOuO/SMIE.