Computer Vision and Pattern Recognition (cs.CV)

Abstract: Recent advances in neural reconstruction using posed image sequences have made remarkable progress. However, due to the lack of depth information, existing volumetric-based techniques simply duplicate 2D image features of the object surface along the entire camera ray. We contend this duplication introduces noise in empty and occluded spaces, posing challenges for producing high-quality 3D geometry. Drawing inspiration from traditional multi-view stereo methods, we propose an end-to-end 3D neural reconstruction framework CVRecon, designed to exploit the rich geometric embedding in the cost volumes to facilitate 3D geometric feature learning. Furthermore, we present Ray-contextual Compensated Cost Volume (RCCV), a novel 3D geometric feature representation that encodes view-dependent information with improved integrity and robustness. Through comprehensive experiments, we demonstrate that our approach significantly improves the reconstruction quality in various metrics and recovers clear fine details of the 3D geometries. Our extensive ablation studies provide insights into the development of effective 3D geometric feature learning schemes. Project page: https://cvrecon.ziyue.cool/

Abstract: Live video commenting is popular on video media platforms, as it can create a chatting atmosphere and provide supplementary information for users while watching videos. Automatically generating live video comments can improve user experience and enable human-like generation for bot chatting. Existing works mostly focus on short video datasets while ignoring other important video types such as long videos like movies. In this work, we collect a new Movie Live Comments (MovieLC) dataset to support research on live video comment generation for long videos. We also propose a knowledge enhanced generation model inspired by the divergent and informative nature of live video comments. Our model adopts a pre-training encoder-decoder framework and incorporates external knowledge. Extensive experiments show that both objective metrics and human evaluation demonstrate the effectiveness of our proposed model. The MovieLC dataset and our code will be released.

Abstract: The proliferation of videos collected during in-the-wild natural settings has pushed the development of effective Video Quality Assessment (VQA) methodologies. Contemporary supervised opinion-driven VQA strategies predominantly hinge on training from expensive human annotations for quality scores, which limited the scale and distribution of VQA datasets and consequently led to unsatisfactory generalization capacity of methods driven by these data. On the other hand, although several handcrafted zero-shot quality indices do not require training from human opinions, they are unable to account for the semantics of videos, rendering them ineffective in comprehending complex authentic distortions (e.g., white balance, exposure) and assessing the quality of semantic content within videos. To address these challenges, we introduce the text-prompted Semantic Affinity Quality Index (SAQI) and its localized version (SAQI-Local) using Contrastive Language-Image Pre-training (CLIP) to ascertain the affinity between textual prompts and visual features, facilitating a comprehensive examination of semantic quality concerns without the reliance on human quality annotations. By amalgamating SAQI with existing low-level metrics, we propose the unified Blind Video Quality Index (BVQI) and its improved version, BVQI-Local, which demonstrates unprecedented performance, surpassing existing zero-shot indices by at least 24\% on all datasets. Moreover, we devise an efficient fine-tuning scheme for BVQI-Local that jointly optimizes text prompts and final fusion weights, resulting in state-of-the-art performance and superior generalization ability in comparison to prevalent opinion-driven VQA methods. We conduct comprehensive analyses to investigate different quality concerns of distinct indices, demonstrating the effectiveness and rationality of our design.

Abstract: People with vocal and hearing disabilities use sign language to express themselves using visual gestures and signs. Although sign language is a solution for communication difficulties faced by deaf people, there are still problems as most of the general population cannot understand this language, creating a communication barrier, especially in places such as banks, airports, supermarkets, etc. [1]. A sign language recognition(SLR) system is a must to solve this problem. The main focus of this model is to develop a real-time word-level sign language recognition system that would translate sign language to text. Much research has been done on ASL(American sign language). Thus, we have worked on ISL(Indian sign language) to cater to the needs of the deaf and hard-of-hearing community of India[2]. In this research, we provide an Indian Sign Language-based Sign Language recognition system. For this analysis, the user must be able to take pictures of hand movements using a web camera, and the system must anticipate and display the name of the taken picture. The acquired image goes through several processing phases, some of which use computer vision techniques, including grayscale conversion, dilatation, and masking. Our model is trained using a convolutional neural network (CNN), which is then utilized to recognize the images. Our best model has a 99% accuracy rate[3].

Abstract: In the field of affective computing, researchers in the community have promoted the performance of models and algorithms by using the complementarity of multimodal information. However, the emergence of more and more modal information makes the development of datasets unable to keep up with the progress of existing modal sensing equipment. Collecting and studying multimodal data is a complex and significant work. In order to supplement the challenge of partial missing of community data. We collected and labeled a new homogeneous multimodal gesture emotion recognition dataset based on the analysis of the existing data sets. This data set complements the defects of homogeneous multimodal data and provides a new research direction for emotion recognition. Moreover, we propose a pseudo dual-flow network based on this dataset, and verify the application potential of this dataset in the affective computing community. The experimental results demonstrate that it is feasible to use the traditional visual information and spiking visual information based on homogeneous multimodal data for visual emotion recognition.The dataset is available at \url{https://github.com/201528014227051/SGED}

Abstract: The success of deep learning is frequently described as the ability to train all parameters of a network on a specific application in an end-to-end fashion. Yet, several design choices on the camera level, including the pixel layout of the sensor, are considered as pre-defined and fixed, and high resolution, regular pixel layouts are considered to be the most generic ones in computer vision and graphics, treating all regions of an image as equally important. While several works have considered non-uniform, \eg, hexagonal or foveated, pixel layouts in hardware and image processing, the layout has not been integrated into the end-to-end learning paradigm so far. In this work, we present the first truly end-to-end trained imaging pipeline that optimizes the size and distribution of pixels on the imaging sensor jointly with the parameters of a given neural network on a specific task. We derive an analytic, differentiable approach for the sensor layout parameterization that allows for task-specific, local varying pixel resolutions. We present two pixel layout parameterization functions: rectangular and curvilinear grid shapes that retain a regular topology. We provide a drop-in module that approximates sensor simulation given existing high-resolution images to directly connect our method with existing deep learning models. We show that network predictions benefit from learnable pixel layouts for two different downstream tasks, classification and semantic segmentation.

Abstract: Losing pets can be highly distressing for pet owners, and finding a lost pet is often challenging and time-consuming. An artificial intelligence-based application can significantly improve the speed and accuracy of finding lost pets. In order to facilitate such an application, this study introduces a contrastive neural network model capable of accurately distinguishing between images of pets. The model was trained on a large dataset of dog images and evaluated through 3-fold cross-validation. Following 350 epochs of training, the model achieved a test accuracy of 90%. Furthermore, overfitting was avoided, as the test accuracy closely matched the training accuracy. Our findings suggest that contrastive neural network models hold promise as a tool for locating lost pets. This paper provides the foundation for a potential web application that allows users to upload images of their missing pets, receiving notifications when matching images are found in the application's image database. This would enable pet owners to quickly and accurately locate lost pets and reunite them with their families.

Abstract: The growing availability of digitized art collections has created the need to manage, analyze and categorize large amounts of data related to abstract concepts, highlighting a demanding problem of computer science and leading to new research perspectives. Advances in artificial intelligence and neural networks provide the right tools for this challenge. The analysis of artworks to extract features useful in certain works is at the heart of the era. In the present work, we approach the problem of painter recognition in a set of digitized paintings, derived from the WikiArt repository, using transfer learning to extract the appropriate features and classical machine learning methods to evaluate the result. Through the testing of various models and their fine tuning we came to the conclusion that RegNet performs better in exporting features, while SVM makes the best classification of images based on the painter with a performance of up to 85%. Also, we introduced a new large dataset for painting recognition task including 62 artists achieving good results.

Abstract: Remote photoplethysmography (rPPG) offers a state-of-the-art, non-contact methodology for estimating human pulse by analyzing facial videos. Despite its potential, rPPG methods can be susceptible to various artifacts, such as noise, occlusions, and other obstructions caused by sunglasses, masks, or even involuntary facial contact, such as individuals inadvertently touching their faces. In this study, we apply image processing transformations to intentionally degrade video quality, mimicking these challenging conditions, and subsequently evaluate the performance of both non-learning and learning-based rPPG methods on the deteriorated data. Our results reveal a significant decrease in accuracy in the presence of these artifacts, prompting us to propose the application of restoration techniques, such as denoising and inpainting, to improve heart-rate estimation outcomes. By addressing these challenging conditions and occlusion artifacts, our approach aims to make rPPG methods more robust and adaptable to real-world situations. To assess the effectiveness of our proposed methods, we undertake comprehensive experiments on three publicly available datasets, encompassing a wide range of scenarios and artifact types. Our findings underscore the potential to construct a robust rPPG system by employing an optimal combination of restoration algorithms and rPPG techniques. Moreover, our study contributes to the advancement of privacy-conscious rPPG methodologies, thereby bolstering the overall utility and impact of this innovative technology in the field of remote heart-rate estimation under realistic and diverse conditions.

Abstract: 3D point cloud semantic segmentation is one of the fundamental tasks for environmental understanding. Although significant progress has been made in recent years, the performance of classes with few examples or few points is still far from satisfactory. In this paper, we propose a novel multi-to-single knowledge distillation framework for the 3D point cloud semantic segmentation task to boost the performance of those hard classes. Instead of fusing all the points of multi-scans directly, only the instances that belong to the previously defined hard classes are fused. To effectively and sufficiently distill valuable knowledge from multi-scans, we leverage a multilevel distillation framework, i.e., feature representation distillation, logit distillation, and affinity distillation. We further develop a novel instance-aware affinity distillation algorithm for capturing high-level structural knowledge to enhance the distillation efficacy for hard classes. Finally, we conduct experiments on the SemanticKITTI dataset, and the results on both the validation and test sets demonstrate that our method yields substantial improvements compared with the baseline method. The code is available at \Url{https://github.com/skyshoumeng/M2SKD}.

Abstract: Labeling LiDAR point clouds for training autonomous driving is extremely expensive and difficult. LiDAR simulation aims at generating realistic LiDAR data with labels for training and verifying self-driving algorithms more efficiently. Recently, Neural Radiance Fields (NeRF) have been proposed for novel view synthesis using implicit reconstruction of 3D scenes. Inspired by this, we present NeRF-LIDAR, a novel LiDAR simulation method that leverages real-world information to generate realistic LIDAR point clouds. Different from existing LiDAR simulators, we use real images and point cloud data collected by self-driving cars to learn the 3D scene representation, point cloud generation and label rendering. We verify the effectiveness of our NeRF-LiDAR by training different 3D segmentation models on the generated LiDAR point clouds. It reveals that the trained models are able to achieve similar accuracy when compared with the same model trained on the real LiDAR data. Besides, the generated data is capable of boosting the accuracy through pre-training which helps reduce the requirements of the real labeled data.

Abstract: Previous methods solve feature matching and pose estimation using a two-stage process by first finding matches and then estimating the pose. As they ignore the geometric relationships between the two tasks, they focus on either improving the quality of matches or filtering potential outliers, leading to limited efficiency or accuracy. In contrast, we propose an iterative matching and pose estimation framework (IMP) leveraging the geometric connections between the two tasks: a few good matches are enough for a roughly accurate pose estimation; a roughly accurate pose can be used to guide the matching by providing geometric constraints. To this end, we implement a geometry-aware recurrent attention-based module which jointly outputs sparse matches and camera poses. Specifically, for each iteration, we first implicitly embed geometric information into the module via a pose-consistency loss, allowing it to predict geometry-aware matches progressively. Second, we introduce an \textbf{e}fficient IMP, called EIMP, to dynamically discard keypoints without potential matches, avoiding redundant updating and significantly reducing the quadratic time complexity of attention computation in transformers. Experiments on YFCC100m, Scannet, and Aachen Day-Night datasets demonstrate that the proposed method outperforms previous approaches in terms of accuracy and efficiency.

Abstract: 3D shape reconstruction typically requires identifying object features or textures in multiple images of a subject. This approach is not viable when the subject is semi-transparent and moving in and out of focus. Here we overcome these challenges by rendering a candidate shape with adaptive blurring and transparency for comparison with the images. We use the microscopic nematode Caenorhabditis elegans as a case study as it freely explores a 3D complex fluid with constantly changing optical properties. We model the slender worm as a 3D curve using an intrinsic parametrisation that naturally admits biologically-informed constraints and regularisation. To account for the changing optics we develop a novel differentiable renderer to construct images from 2D projections and compare against raw images to generate a pixel-wise error to jointly update the curve, camera and renderer parameters using gradient descent. The method is robust to interference such as bubbles and dirt trapped in the fluid, stays consistent through complex sequences of postures, recovers reliable estimates from blurry images and provides a significant improvement on previous attempts to track C. elegans in 3D. Our results demonstrate the potential of direct approaches to shape estimation in complex physical environments in the absence of ground-truth data.

Abstract: Visual localization is a fundamental task for various applications including autonomous driving and robotics. Prior methods focus on extracting large amounts of often redundant locally reliable features, resulting in limited efficiency and accuracy, especially in large-scale environments under challenging conditions. Instead, we propose to extract globally reliable features by implicitly embedding high-level semantics into both the detection and description processes. Specifically, our semantic-aware detector is able to detect keypoints from reliable regions (e.g. building, traffic lane) and suppress unreliable areas (e.g. sky, car) implicitly instead of relying on explicit semantic labels. This boosts the accuracy of keypoint matching by reducing the number of features sensitive to appearance changes and avoiding the need of additional segmentation networks at test time. Moreover, our descriptors are augmented with semantics and have stronger discriminative ability, providing more inliers at test time. Particularly, experiments on long-term large-scale visual localization Aachen Day-Night and RobotCar-Seasons datasets demonstrate that our model outperforms previous local features and gives competitive accuracy to advanced matchers but is about 2 and 3 times faster when using 2k and 4k keypoints, respectively.

Abstract: Weather recognition is an essential support for many practical life applications, including traffic safety, environment, and meteorology. However, many existing related works cannot comprehensively describe weather conditions due to their complex co-occurrence dependencies. This paper proposes a novel multi-label weather recognition model considering these dependencies. The proposed model called MASK-Convolutional Neural Network-Transformer (MASK-CT) is based on the Transformer, the convolutional process, and the MASK mechanism. The model employs multiple convolutional layers to extract features from weather images and a Transformer encoder to calculate the probability of each weather condition based on the extracted features. To improve the generalization ability of MASK-CT, a MASK mechanism is used during the training phase. The effect of the MASK mechanism is explored and discussed. The Mask mechanism randomly withholds some information from one-pair training instances (one image and its corresponding label). There are two types of MASK methods. Specifically, MASK-I is designed and deployed on the image before feeding it into the weather feature extractor and MASK-II is applied to the image label. The Transformer encoder is then utilized on the randomly masked image features and labels. The experimental results from various real-world weather recognition datasets demonstrate that the proposed MASK-CT model outperforms state-of-the-art methods. Furthermore, the high-speed dynamic real-time weather recognition capability of the MASK-CT is evaluated.

Abstract: In this paper, we present a dense hybrid proposal modulation (DHPM) method for lane detection. Most existing methods perform sparse supervision on a subset of high-scoring proposals, while other proposals fail to obtain effective shape and location guidance, resulting in poor overall quality. To address this, we densely modulate all proposals to generate topologically and spatially high-quality lane predictions with discriminative representations. Specifically, we first ensure that lane proposals are physically meaningful by applying single-lane shape and location constraints. Benefitting from the proposed proposal-to-label matching algorithm, we assign each proposal a target ground truth lane to efficiently learn from spatial layout priors. To enhance the generalization and model the inter-proposal relations, we diversify the shape difference of proposals matching the same ground-truth lane. In addition to the shape and location constraints, we design a quality-aware classification loss to adaptively supervise each positive proposal so that the discriminative power can be further boosted. Our DHPM achieves very competitive performances on four popular benchmark datasets. Moreover, we consistently outperform the baseline model on most metrics without introducing new parameters and reducing inference speed.

Abstract: Building 3D scene graphs has recently emerged as a topic in scene representation for several embodied AI applications to represent the world in a structured and rich manner. With their increased use in solving downstream tasks (eg, navigation and room rearrangement), can we leverage and recycle them for creating 3D maps of environments, a pivotal step in agent operation? We focus on the fundamental problem of aligning pairs of 3D scene graphs whose overlap can range from zero to partial and can contain arbitrary changes. We propose SGAligner, the first method for aligning pairs of 3D scene graphs that is robust to in-the-wild scenarios (ie, unknown overlap -- if any -- and changes in the environment). We get inspired by multi-modality knowledge graphs and use contrastive learning to learn a joint, multi-modal embedding space. We evaluate on the 3RScan dataset and further showcase that our method can be used for estimating the transformation between pairs of 3D scenes. Since benchmarks for these tasks are missing, we create them on this dataset. The code, benchmark, and trained models are available on the project website.

Abstract: Model merging (e.g., via interpolation or task arithmetic) fuses multiple models trained on different tasks to generate a multi-task solution. The technique has been proven successful in previous studies, where the models are trained on similar tasks and with the same initialization. In this paper, we expand on this concept to a multimodal setup by merging transformers trained on different modalities. Furthermore, we conduct our study for a novel goal where we can merge vision, language, and cross-modal transformers of a modality-specific architecture to create a parameter-efficient modality-agnostic architecture. Through comprehensive experiments, we systematically investigate the key factors impacting model performance after merging, including initialization, merging mechanisms, and model architectures. Our analysis leads to an effective training recipe for matching the performance of the modality-agnostic baseline (i.e. pre-trained from scratch) via model merging. Our code is available at: https://github.com/ylsung/vl-merging

Abstract: Background: Physiological tremor is defined as an involuntary and rhythmic shaking. Tremor of the hand is a key symptom of multiple neurological diseases, and its frequency and amplitude differs according to both disease type and disease progression. In routine clinical practice, tremor frequency and amplitude are assessed by expert rating using a 0 to 4 integer scale. Such ratings are subjective and have poor inter-rater reliability. There is thus a clinical need for a practical and accurate method for objectively assessing hand tremor. Objective: to develop a proof of principle method to measure hand tremor amplitude from smartphone videos. Methods: We created a computer vision pipeline that automatically extracts salient points on the hand and produces a 1-D time series of movement due to tremor, in pixels. Using the smartphones' depth measurement, we convert this measure into real distance units. We assessed the accuracy of the method using 60 videos of simulated tremor of different amplitudes from two healthy adults. Videos were taken at distances of 50, 75 and 100 cm between hand and camera. The participants had skin tone II and VI on the Fitzpatrick scale. We compared our method to a gold-standard measurement from a slide rule. Bland-Altman methods agreement analysis indicated a bias of 0.04 cm and 95% limits of agreement from -1.27 to 1.20 cm. Furthermore, we qualitatively observed that the method was robust to differences in skin tone and limited occlusion, such as a band-aid affixed to the participant's hand. Clinical relevance: We have demonstrated how tremor amplitude can be measured from smartphone videos. In conjunction with tremor frequency, this approach could be used to help diagnose and monitor neurological diseases

Abstract: Trends and opinion mining in social media increasingly focus on novel interactions involving visual media, like images and short videos, in addition to text. In this work, we tackle the problem of visual sentiment analysis of social media images -- specifically, the prediction of image sentiment polarity. While previous work relied on manually labeled training sets, we propose an automated approach for building sentiment polarity classifiers based on a cross-modal distillation paradigm; starting from scraped multimodal (text + images) data, we train a student model on the visual modality based on the outputs of a textual teacher model that analyses the sentiment of the corresponding textual modality. We applied our method to randomly collected images crawled from Twitter over three months and produced, after automatic cleaning, a weakly-labeled dataset of $\sim$1.5 million images. Despite exploiting noisy labeled samples, our training pipeline produces classifiers showing strong generalization capabilities and outperforming the current state of the art on five manually labeled benchmarks for image sentiment polarity prediction.

Abstract: A road is the skeleton of a city and is a fundamental and important geographical component. Currently, many countries have built geo-information databases and gathered large amounts of geographic data. However, with the extensive construction of infrastructure and rapid expansion of cities, automatic updating of road data is imperative to maintain the high quality of current basic geographic information. However, obtaining bi-phase images for the same area is difficult, and complex post-processing methods are required to update the existing databases.To solve these problems, we proposed a road detection method based on semi-supervised learning (SRUNet) specifically for road-updating applications; in this approach, historical road information was fused with the latest images to directly obtain the latest state of the road.Considering that the texture of a road is complex, a multi-branch network, named the Map Encoding Branch (MEB) was proposed for representation learning, where the Boundary Enhancement Module (BEM) was used to improve the accuracy of boundary prediction, and the Residual Refinement Module (RRM) was used to optimize the prediction results. Further, to fully utilize the limited amount of label information and to enhance the prediction accuracy on unlabeled images, we utilized the mean teacher framework as the basic semi-supervised learning framework and introduced Regional Contrast (ReCo) in our work to improve the model capacity for distinguishing between the characteristics of roads and background elements.We applied our method to two datasets. Our model can effectively improve the performance of a model with fewer labels. Overall, the proposed SRUNet can provide stable, up-to-date, and reliable prediction results for a wide range of road renewal tasks.

Abstract: SplitFed Learning, a combination of Federated and Split Learning (FL and SL), is one of the most recent developments in the decentralized machine learning domain. In SplitFed learning, a model is trained by clients and a server collaboratively. For image segmentation, labels are created at each client independently and, therefore, are subject to clients' bias, inaccuracies, and inconsistencies. In this paper, we propose a data quality-based adaptive averaging strategy for SplitFed learning, called QA-SplitFed, to cope with the variation of annotated ground truth (GT) quality over multiple clients. The proposed method is compared against five state-of-the-art model averaging methods on the task of learning human embryo image segmentation. Our experiments show that all five baseline methods fail to maintain accuracy as the number of corrupted clients increases. QA-SplitFed, however, copes effectively with corruption as long as there is at least one uncorrupted client.

Abstract: When applied to Image-to-text models, interpretability methods often provide token-by-token explanations namely, they compute a visual explanation for each token of the generated sequence. Those explanations are expensive to compute and unable to comprehensively explain the model's output. Therefore, these models often require some sort of approximation that eventually leads to misleading explanations. We develop a framework based on SHAP, that allows for generating comprehensive, meaningful explanations leveraging the meaning representation of the output sequence as a whole. Moreover, by exploiting semantic priors in the visual backbone, we extract an arbitrary number of features that allows the efficient computation of Shapley values on large-scale models, generating at the same time highly meaningful visual explanations. We demonstrate that our method generates semantically more expressive explanations than traditional methods at a lower compute cost and that it can be generalized over other explainability methods.

Abstract: How to efficiently transform large language models (LLMs) into instruction followers is recently a popular research direction, while training LLM for multi-modal reasoning remains less explored. Although the recent LLaMA-Adapter demonstrates the potential to handle visual inputs with LLMs, it still cannot generalize well to open-ended visual instructions and lags behind GPT-4. In this paper, we present LLaMA-Adapter V2, a parameter-efficient visual instruction model. Specifically, we first augment LLaMA-Adapter by unlocking more learnable parameters (e.g., norm, bias and scale), which distribute the instruction-following ability across the entire LLaMA model besides adapters. Secondly, we propose an early fusion strategy to feed visual tokens only into the early LLM layers, contributing to better visual knowledge incorporation. Thirdly, a joint training paradigm of image-text pairs and instruction-following data is introduced by optimizing disjoint groups of learnable parameters. This strategy effectively alleviates the interference between the two tasks of image-text alignment and instruction following and achieves strong multi-modal reasoning with only a small-scale image-text and instruction dataset. During inference, we incorporate additional expert models (e.g. captioning/OCR systems) into LLaMA-Adapter to further enhance its image understanding capability without incurring training costs. Compared to the original LLaMA-Adapter, our LLaMA-Adapter V2 can perform open-ended multi-modal instructions by merely introducing 14M parameters over LLaMA. The newly designed framework also exhibits stronger language-only instruction-following capabilities and even excels in chat interactions. Our code and models are available at https://github.com/ZrrSkywalker/LLaMA-Adapter.