Vision-language models (VLMs) have shown impressive performance in substantial downstream multi-modal tasks. However, only comparing the fine-tuned performance on downstream tasks leads to the poor interpretability of VLMs, which is adverse to their future improvement. Several prior works have identified this issue and used various probing methods under a zero-shot setting to detect VLMs' limitations, but they all examine VLMs using general datasets instead of specialized ones. In practical applications, VLMs are usually applied to specific scenarios, such as e-commerce and news fields, so the generalization of VLMs in specific domains should be given more attention. In this paper, we comprehensively investigate the capabilities of popular VLMs in a specific field, the food domain. To this end, we build a food caption dataset, Food-500 Cap, which contains 24,700 food images with 494 categories. Each image is accompanied by a detailed caption, including fine-grained attributes of food, such as the ingredient, shape, and color. We also provide a culinary culture taxonomy that classifies each food category based on its geographic origin in order to better analyze the performance differences of VLM in different regions. Experiments on our proposed datasets demonstrate that popular VLMs underperform in the food domain compared with their performance in the general domain. Furthermore, our research reveals severe bias in VLMs' ability to handle food items from different geographic regions. We adopt diverse probing methods and evaluate nine VLMs belonging to different architectures to verify the aforementioned observations. We hope that our study will bring researchers' attention to VLM's limitations when applying them to the domain of food or culinary cultures, and spur further investigations to address this issue.
论文检索
输入标题、作者或关键词，从 12,319 篇学术成果中精准定位
Visual Commonsense Reasoning (VCR) calls for explanatory reasoning behind question answering over visual scenes. To achieve this goal, a model is required to provide an acceptable rationale as the reason for the predicted answers. Progress on the benchmark dataset stems largely from the recent advancement of Vision-Language Transformers (VL Transformers). These models are first pre-trained on some generic large-scale vision-text datasets, and then the learned representations are transferred to the downstream VCR task. Despite their attractive performance, this paper posits that the VL Transformers do not exhibit visual commonsense, which is the key to VCR. In particular, our empirical results pinpoint several shortcomings of existing VL Transformers: small gains from pre-training, unexpected language bias, limited model architecture for the two inseparable sub-tasks, and neglect of the important object-tag correlation. With these findings, we tentatively suggest some future directions from the aspect of dataset, evaluation metric, and training tricks. We believe this work could make researchers revisit the intuition and goals of VCR, and thus help tackle the remaining challenges in visual reasoning.
Although the attention mechanism in transformers has proven successful in image-text retrieval tasks, most transformer models suffer from a large number of parameters. Inspired by brain circuits that process information with recurrent connected neurons, we propose a novel Reservoir Computing Transformer Reasoning Network (RCTRN) for image-text retrieval. The proposed RCTRN employs a two-step strategy to focus on feature representation and data distribution of different modalities respectively. Specifically, we send visual and textual features through a unified meshed reasoning module, which encodes multi-level feature relationships with prior knowledge and aggregates the complementary outputs in a more effective way. The reservoir reasoning network is proposed to optimize memory connections between features at different stages and address the data distribution mismatch problem introduced by the unified scheme. To investigate the significance of the low power dissipation and low bandwidth characteristics of RRN in practical scenarios, we deployed the model in the wireless transmission system, demonstrating that RRN's optimization of data structures also has a certain robustness against channel noise. Extensive experiments on two benchmark datasets, Flickr30K and MS-COCO, demonstrate the superiority of RCTRN in terms of performance and low-power dissipation compared to state-of-the-art baselines.
Hierarchical Prompt Learning Using CLIP for Multi-label Classification with Single Positive Labels
PDF ↗Collecting full annotations to construct multi-label datasets is difficult and labor-consuming. As an effective solution to relieve the annotation burden, single positive multi-label learning (SPML) draws increasing attention from both academia and industry. It only annotates each image with one positive label, leaving other labels unobserved. Therefore, existing methods strive to explore the cue of unobserved labels to compensate for the insufficiency of label supervision. Though achieving promising performance, they generally consider labels independently, leaving out the inherent hierarchical semantic relationship among labels which reveals that labels can be clustered into groups. In this paper, we propose a hierarchical prompt learning method with a novel Hierarchical Semantic Prompt Network (HSPNet) to harness such hierarchical semantic relationships using a large-scale pretrained vision and language model, i.e., CLIP, for SPML. We first introduce a Hierarchical Conditional Prompt (HCP) strategy to grasp the hierarchical label-group dependency. Then we equip a Hierarchical Graph Convolutional Network (HGCN) to capture the high-order inter-label and inter-group dependencies. Comprehensive experiments and analyses on several benchmark datasets show that our method significantly outperforms the state-of-the-art methods, well demonstrating its superiority and effectiveness. Our code will be available at https://github.com/jameslahm/HSPNet.
PoSynDA: Multi-Hypothesis Pose Synthesis Domain Adaptation for Robust 3D Human Pose Estimation
PDF ↗The current 3D human pose estimators face challenges in adapting to new datasets due to the scarcity of 2D-3D pose pairs in target domain training sets. We present the Multi-Hypothesis Pose Synthesis Domain Adaptation (PoSynDA) framework to overcome this issue without extensive target domain annotation. Utilizing a diffusion-centric structure, PoSynDA simulates the 3D pose distribution in the target domain, filling the data diversity gap. By incorporating a multi-hypothesis network, it creates diverse pose hypotheses and aligns them with the target domain. Target-specific source augmentation obtains the target domain distribution data from the source domain by decoupling the scale and position parameters. The teacher-student paradigm and low-rank adaptation further refine the process. PoSynDA demonstrates competitive performance on benchmarks, such as Human3.6M, MPI-INF-3DHP, and 3DPW, even comparable with the target-trained MixSTE model. This work paves the way for the practical application of 3D human pose estimation1. The source code is available at https://github.com/hbing-l/PoSynDA.
Unsupervised salient object detection aims to detect salient objects without using supervision signals eliminating the tedious task of manually labeling salient objects. To improve training efficiency, end-to-end methods for USOD have been proposed as a promising alternative. However, current solutions rely heavily on noisy handcraft labels and fail to mine rich semantic information from deep features. In this paper, we propose a self-supervised end-to-end salient object detection framework via top-down context. Specifically, motivated by contrastive learning, we exploit the self-localization from the deepest feature to construct the location maps which are then leveraged to learn the most instructive segmentation guidance. Further considering the lack of detailed information in deepest features, we exploit the detail-boosting refiner module to enrich the location labels with details. Moreover, we observe that due to lack of supervision, current unsupervised saliency models tend to detect non-salient objects that are salient in some other samples of corresponding scenarios. To address this widespread issue, we design a novel Unsupervised Non-Salient Suppression (UNSS) method developing the ability to ignore non-salient objects. Extensive experiments on benchmark datasets demonstrate that our method achieves leading performance among the recent end-to-end methods and most of the multi-stage solutions. The code is available.
In this paper, we introduce a novel approach to address the task of synthesizing speech from silent videos of any in-the-wild speaker solely based on lip movements. The traditional approach of directly generating speech from lip videos faces the challenge of not being able to learn a robust language model from speech alone, resulting in unsatisfactory outcomes. To overcome this issue, we propose incorporating noisy text supervision using a state-of-the-art lip-to-text network that instills language information into our model. The noisy text is generated using a pre-trained lip-to-text model, enabling our approach to work without text annotations during inference. We design a visual text-to-speech network that utilizes the visual stream to generate accurate speech, which is in-sync with the silent input video. We perform extensive experiments and ablation studies, demonstrating our approach's superiority over the current state-of-the-art methods on various benchmark datasets. Further, we demonstrate an essential practical application of our method in assistive technology by generating speech for an ALS patient who has lost the voice but can make mouth movements. Our demo video, code, and additional details can be found at http://cvit.iiit.ac.in/research/projects/cvit-projects/ms-l2s-itw.
Contemporary news reporting increasingly features multimedia content, motivating research on multimedia event extraction. However, the task lacks annotated multimodal training data and artificially generated training data suffer from the distribution shift from the real-world data. In this paper, we propose Cross-modality Augmented Multimedia Event Learning (CAMEL), which successfully utilizes artificially generated multimodal training data and achieves state-of-the-art performance. Conditioned on unimodal training data, we generate multimodal training data using off-the-shelf image generators like Stable Diffusion [45] and image captioners like BLIP [24]. After that, we train the network on the resultant multimodal datasets. In order to learn robust features that are effective across domains, we devise an iterative and gradual training strategy. Substantial experiments show that CAMEL surpasses state-of-the-art (SOTA) baselines on the M2E2 benchmark. On multimedia events in particular, we outperform the prior SOTA by 4.2% F1 on event mention identification and by 9.8% F1 on argument identification, which demonstrates that CAMEL learns synergistic representations from the two modalities. Our work demonstrates a recipe to unleash the power of synthetic training data in structured prediction.
Spectral modeling plays a critical role in denoising hyperspectral images (HSIs), with recent approaches leveraging well-designed network architectures to extract spectral contexts for noise removal. However, these approaches overlook a striking finding: the presence of spectral differences in noisy contexts can pose challenges for the denoising network during the restoration process of each band in the HSI. We attribute this to the varying levels of spectral difference between different bands and the unknown distribution of various noises. These factors can make it difficult for the network to capture consistent features, ultimately leading to suboptimal solutions. We propose a novel concept termed 'spectral displacement,' which views spectral differences as pixel motion displacement along the spectral domain. To eliminate the effect of spectral displacement, we introduce a potential solution: spectral alignment. This approach can increase the mutual information between different spectral bands and enhance the effectiveness of denoising. We then present the Spectral Alignment Recurrent Network (SARN) for efficient and effective displacement estimation and pixel-level alignment between neighboring bands. SARN can serve as a general plug-in for HSI backbones without requiring any model-specific design. Experimental results on several benchmark datasets confirm the effectiveness and superiority of our concept and network. The source code will be available at https://github.com/MIV-XJTU/SARN.
Human-Object Interaction (HOI) detection aims to understand the interactions between humans and objects, which plays a curtail role in high-level semantic understanding tasks. However, most works pursue designing better architectures to learn overall features more efficiently, while ignoring the long-tail nature of interaction-object pair categories. In this paper, we propose to alleviate the impact of such an unbalanced distribution via Virtual Image Leaning (VIL). Firstly, a novel label-to-image approach, Multiple Steps Image Creation (MUSIC), is proposed to create a high-quality dataset that has a consistent distribution with real images. In this stage, virtual images are generated based on prompts with specific characterizations and selected by multi-filtering processes. Secondly, we use both virtual and real images to train the model with the teacher-student framework. Considering the initial labels of some virtual images are inaccurate and inadequate, we devise an Adaptive Matching-and-Filtering (AMF) module to construct pseudo-labels. Our method is independent of the internal structure of HOI detectors, so it can be combined with off-the-shelf methods by training merely 10 additional epochs. With the assistance of our method, multiple methods obtain significant improvements, and new state-of-the-art results are achieved on two benchmarks.
With the increasing volume of healthcare data, automated International Classification of Diseases (ICD) has become increasingly relevant and is frequently regarded as a medical multi-label prediction problem. Current methods struggle to accurately classify medical diagnosis texts that represent deep and sparse categories. Unlike these works that model the label with code hierarchy or description for label prediction, we argue that the label generation with structural information can provide more comprehensive knowledge based on the observation that label synonyms and parent-child relationships in vary from their context in clinical contexts. In this study, we introduce \tool, a heterogeneous graph model with improved attention for automated ICD coding. Notably, our approach represents the model to consider this task as a labelled graph generation problem. Our enhanced attention mechanism boosts the model's capacity to learn from multi-relational heterogeneous graph representations. Additionally, we propose a discriminator for labelled graphs (LG) that computes the reward for each ICD code in the labelled graph generator. Our experimental findings demonstrate that our proposed model significantly outperforms all existing strong baseline methods and attains the best performance on three benchmark datasets.
Image-text retrieval is a fundamental branch in cross-modal retrieval. The core is to explore the semantic correspondence to align relevant image-text pairs. Some existing methods rely on global semantics and co-occurrence frequency to design knowledge introduction patterns for consistent representations. However, they lack flexibility due to the limitations of fixed information and empirical feedback. To address these issues, we develop an External Knowledge Dynamic Modeling~(EKDM) architecture based on the filtering mechanism, which dynamically explores different knowledge towards varied image-text pairs. Specially, we first capture abundant concepts and relationships from external knowledge to construct visual and textual corpus sets. Then, we progressively explores concepts related to images and texts by dynamic global representations. To endow the model with the capability of relationship decision, we integrate the variable spatial locations between objects for association exploration. Since the filtering mechanism is conditioned on dynamic semantics and variable spatial locations, our model can dynamically model different knowledge for different image-text pairs. Extensive experimental results on two benchmark datasets demonstrate the effectiveness of our proposed method.
Contrast-augmented Diffusion Model with Fine-grained Sequence Alignment for Markup-to-Image Generation
PDF ↗The recently rising markup-to-image generation poses greater challenges as compared to natural image generation, due to its low tolerance for errors as well as the complex sequence and context correlations between markup and rendered image. This paper proposes a novel model named "Contrast-augmented Diffusion Model with Fine-grained Sequence Alignment'' (FSA-CDM), which introduces contrastive positive/negative samples into the diffusion model to boost performance for markup-to-image generation. Technically, we design a fine-grained cross-modal alignment module to well explore the sequence similarity between the two modalities for learning robust feature representations. To improve the generalization ability, we propose a contrast-augmented diffusion model to explicitly explore positive and negative samples by maximizing a novel contrastive variational objective, which is mathematically inferred to provide a tighter bound for the model's optimization. Moreover, the context-aware cross attention module is developed to capture the contextual information within markup language during the denoising process, yielding better noise prediction results. Extensive experiments are conducted on four benchmark datasets from different domains, and the experimental results demonstrate the effectiveness of the proposed components in FSA-CDM, significantly exceeding state-of-the-art performance by about 2% ~ 12% DTW improvements.
As one of the core video semantic understanding tasks, Video Semantic Role Labeling (VidSRL) aims to detect the salient events from given videos, by recognizing the predict-argument event structures and the interrelationships between events. While recent endeavors have put forth methods for VidSRL, they can be mostly subject to two key drawbacks, including the lack of fine-grained spatial scene perception and the insufficiently modeling of video temporality. Towards this end, this work explores a novel holistic spatio-temporal scene graph (namely HostSG) representation based on the existing dynamic scene graph structures, which well model both the fine-grained spatial semantics and temporal dynamics of videos for VidSRL. Built upon the HostSG, we present a nichetargeting VidSRL framework. A scene-event mapping mechanism is first designed to bridge the gap between the underlying scene structure and the high-level event semantic structure, resulting in an overall hierarchical scene-event (termed ICE) graph structure. We further perform iterative structure refinement to optimize the ICE graph, e.g., filtering noisy branches and newly building informative connections, such that the overall structure representation can best coincide with end task demand. Finally, three subtask predictions of VidSRL are jointly decoded, where the end-to-end paradigm effectively avoids error propagation. On the benchmark dataset, our framework boosts significantly over the current best-performing model. Further analyses are shown for a better understanding of the advances of our methods. Our HostSG representation shows greater potential to facilitate a broader range of other video understanding tasks.
Recent studies of two-view correspondence learning usually establish an end-to-end network to jointly predict correspondence reliability and relative pose. We improve such a framework from two aspects. First, we propose a Local Feature Consensus (LFC) plugin block to augment the features of existing models. Given a correspondence feature, the block augments its neighboring features with mutual neighborhood consensus and aggregates them to produce an enhanced feature. As inliers obey a uniform cross-view transformation and share more consistent learned features than outliers, feature consensus strengthens inlier correlation and suppresses outlier distraction, which makes output features more discriminative for classifying inliers/outliers. Second, existing approaches supervise network training with the ground truth correspondences and essential matrix projecting one image to the other for an input image pair, without considering the information from the reverse mapping. We extend existing models to a Siamese network with a reciprocal loss that exploits the supervision of mutual projection, which considerably promotes the matching performance without introducing additional model parameters. Building upon MSA-Net [30], we implement the two proposals and experimentally achieve state-of-the-art performance on benchmark datasets.
Brain-inspired Spiking Neural Networks (SNNs) have the characteristics of event-driven and high energy-efficient, which are different from traditional Artificial Neural Networks (ANNs) when deployed on edge devices such as neuromorphic chips. Most previous work focuses on SNNs training strategies to improve model performance and brings larger and deeper network architectures. It's difficult to deploy these complex networks on resource-limited edge devices directly. To meet such demand, people compress SNNs very cautiously to balance the performance and the computation efficiency. Existing compression methods either iteratively pruned SNNs using weights norm magnitude or formulated the problem as a sparse learning optimization. We propose an improved end-to-end Minimax optimization method for this sparse learning problem to better balance the model performance and the computation efficiency. We also demonstrate that jointly applying compression and finetuning on SNNs is better than sequentially, especially for extreme compression ratios. The compressed SNN models achieved state-of-the-art (SOTA) performance on various benchmark datasets and architectures. Our code is available athttps://github.com/chenjallen/Resource-Constrained-Compression-on-SNN .
Frame-supervised language-driven action localization aims to localize action boundaries in untrimmed videos corresponding to the input natural language query, with only a single frame annotation within the target action in training. This task is challenging due to the absence of complete and accurate annotation of action boundaries, hindering visual-language alignment and action boundary prediction. To address this challenge, we propose a novel method that introduces distribution functions to model both the probability of action frame and that of boundary frame. Specifically, we assign each video frame the probability of being the action frame based on the estimated shape parameters of the distribution function, serving as a foreground pseudo-label that guides cross-modal feature learning. Moreover, we model the probabilities of start frame and end frame of the target action using different distribution functions, and then estimate the probability of each action candidate being a positive candidate based on its start and end boundaries, which facilitates predicting action boundaries by exploring more positive terms in training. Experiments on two benchmark datasets demonstrate that our method outperforms existing methods, achieving a gain of more than 10% of R1@ 0.5 on the challenging TACoS dataset. These results emphasize the significance of generating pseudo labels with appropriate probabilities via distribution functions to address the challenge of frame-supervised language-driven action localization.
A video storyboard is a roadmap for video creation which consists of shot-by-shot images to visualize key plots in a text synopsis. Creating video storyboards, however, remains challenging which not only requires cross-modal association between high-level texts and images but also demands long-term reasoning to make transitions smooth across shots. In this paper, we propose a new task called Text synopsis to Video Storyboard (TeViS) which aims to retrieve an ordered sequence of images as the video storyboard to visualize the text synopsis. We construct a MovieNet-TeViS dataset based on the public MovieNet dataset [17]. It contains 10K text synopses each paired with keyframes manually selected from corresponding movies by considering both relevance and cinematic coherence. To benchmark the task, we present strong CLIP-based baselines and a novel VQ-Trans model. VQ-Trans first encodes text synopsis and images into a joint embedding space and uses vector quantization (VQ) to improve the visual representation. Then, it auto-regressively generates a sequence of visual features for retrieval and ordering. Experimental results demonstrate that VQ-Trans significantly outperforms prior methods and the CLIP-based baselines. Nevertheless, there is still a large gap compared to human performance suggesting room for promising future work. The code and data are available at: https://ruc-aimind.github.io/projects/TeViS/
Deep Neural Networks suffer significant performance degeneration when noisy labels corrupt latent data representations. Previous work has attempted to alleviate this problem by exploiting contrastive learning, the pair building of which is critical. However, existing methods either conduct sample-level processes and then use the resultant subset to construct pairs or directly perform pair-level selecting using a fixed threshold, both leading to sub-optimal pairing and subsequent representation learning. To address this issue, we propose a novel adaptive contrastive learning method (ACL) working at the pair level to select contrastive pairs adaptively. Specifically, we consider the model's learning status to adjust the confidence threshold in a self-adaptive manner instead of fixing it. Then, towards the ineffectiveness of the thresholding method on unconfident pairs, we automatically apply instance-specific temperature to boost the confidence of accurately-predicted samples and their pairs. We further introduce temporal cross-ensembling to handle the impact of noisy labels on model predictions. As a result, diverse pairs are correctly selected for contrastive learning to induce discriminative representations robust to various types of label noise. Extensive experimental results on several standard benchmarks and real-world datasets indicate the superiority of ACL, especially in extremely noisy scenarios.
HSVLT: Hierarchical Scale-Aware Vision-Language Transformer for Multi-Label Image Classification
PDF ↗The task of multi-label image classification involves recognizing multiple objects within a single image. Considering both valuable semantic information contained in the labels and essential visual features presented in the image, tight visual-linguistic interactions play a vital role in improving classification performance. Moreover, given the potential variance in object size and appearance within a single image, attention to features of different scales can help to discover possible objects in the image. Recently, Transformer-based methods have achieved great success in multi-label image classification by leveraging the advantage of modeling long-range dependencies, but they have several limitations. Firstly, existing methods treat visual feature extraction and cross-modal fusion as separate steps, resulting in insufficient visual-linguistic alignment in the joint semantic space. Additionally, they only extract visual features and perform cross-modal fusion at a single scale, neglecting objects with different characteristics. To address these issues, we propose a Hierarchical Scale-Aware Vision-Language Transformer (HSVLT) with two appealing designs: (1)A hierarchical multi-scale architecture that involves a Cross-Scale Aggregation module, which leverages joint multi-modal features extracted from multiple scales to recognize objects of varying sizes and appearances in images. (2)Interactive Visual-Linguistic Attention, a novel attention mechanism module that tightly integrates cross-modal interaction, enabling the joint updating of visual, linguistic and multi-modal features. We have evaluated our method on three benchmark datasets. The experimental results demonstrate that HSVLT surpasses state-of-the-art methods with lower computational cost.