Composed Image Retrieval (CoIR) has recently gained popularity as a task that considers both text and image queries together, to search for relevant images in a database. Most CoIR approaches require manually annotated datasets, comprising image-text-image triplets, where the text describes a modification from the query image to the target image. However, manual curation of CoIR triplets is expensive and prevents scalability. In this work, we instead propose a scalable automatic dataset creation methodology that generates triplets given video-caption pairs, while also expanding the scope of the task to include composed video retrieval (CoVR). To this end, we mine paired videos with a similar caption from a large database, and leverage a large language model to generate the corresponding modification text. Applying this methodology to the extensive WebVid2M collection, we automatically construct our WebVid-CoVR dataset, resulting in 1.6 million triplets. Moreover, we introduce a new benchmark for CoVR with a manually annotated evaluation set, along with baseline results. Our experiments further demonstrate that training a CoVR model on our dataset effectively transfers to CoIR, leading to improved state-of-the-art performance in the zero-shot setup on both the CIRR and FashionIQ benchmarks. Our code, datasets, and models are publicly available at https://imagine.enpc.fr/~ventural/covr.
论文检索
输入标题、作者或关键词，从 2,866 篇学术成果中精准定位
The synthesis of human motion has traditionally been addressed through task-dependent models that focus on specific challenges, such as predicting future motions or filling in intermediate poses conditioned on known key-poses. In this paper, we present a novel task-independent model called UNIMASK-M, which can effectively address these challenges using a unified architecture. Our model obtains comparable or better performance than the state-of-the-art in each field. Inspired by Vision Transformers (ViTs), our UNIMASK-M model decomposes a human pose into body parts to leverage the spatio-temporal relationships existing in human motion. Moreover, we reformulate various pose-conditioned motion synthesis tasks as a reconstruction problem with different masking patterns given as input. By explicitly informing our model about the masked joints, our UNIMASK-M becomes more robust to occlusions. Experimental results show that our model successfully forecasts human motion on the Human3.6M dataset while achieving state-of-the-art results in motion inbetweening on the LaFAN1 dataset for long transition periods.
We contribute to the sparsely populated area of unsupervised deep graph matching with application to keypoint matching in images. Contrary to the standard supervised approach, our method does not require ground truth correspondences between keypoint pairs. Instead, it is self-supervised by enforcing consistency of matchings between images of the same object category. As the matching and the consistency loss are discrete, their derivatives cannot be straightforwardly used for learning. We address this issue in a principled way by building our method upon the recent results on black-box differentiation of combinatorial solvers. This makes our method exceptionally flexible, as it is compatible with arbitrary network architectures and combinatorial solvers. Our experimental evaluation suggests that our technique sets a new state-of-the-art for unsupervised graph matching.
In this paper, we propose a novel transformer-based end-to-end real-time vanishing point detection method, which is named Vanishing Point TRansformer (VPTR). The proposed method can directly regress the locations of vanishing points from given images. To achieve this goal, we pose vanishing point detection as a point object detection task on the Gaussian hemisphere with region division. Considering low-level features always provide more geometric information which can contribute to accurate vanishing point prediction, we propose a clear architecture where vanishing point queries in the decoder can directly gather multi-level features from CNN backbone with deformable attention in VPTR. Our method does not rely on line detection or Manhattan world assumption, which makes it more flexible to use. VPTR runs at an inferring speed of 140 FPS on one NVIDIA 3090 card. Experimental results on synthetic and real-world datasets demonstrate that our method can be used in both natural and structural scenes, and is superior to other state-of-the-art methods on the balance of accuracy and efficiency.
Adversarial training (AT) is an effective defense method against gradient-based attacks to enhance the robustness of neural networks. Among them, single-step AT has emerged as a hotspot topic due to its simplicity and efficiency, requiring only one gradient propagation in generating adversarial examples. Nonetheless, the problem of catastrophic overfitting (CO) that causes training collapse remains poorly understood, and there exists a gap between the robust accuracy achieved through single- and multi-step AT. In this paper, we present a surprising finding that the taxonomy of adversarial examples reveals the truth of CO. Based on this conclusion, we propose taxonomy driven fast adversarial training (TDAT) which jointly optimizes learning objective, loss function, and initialization method, thereby can be regarded as a new paradigm of single-step AT. Compared with other fast AT methods, TDAT can boost the robustness of neural networks, alleviate the influence of misclassified examples, and prevent CO during the training process while requiring almost no additional computational and memory resources. Our method achieves robust accuracy improvement of 1.59%, 1.62%, 0.71%, and 1.26% on CIFAR-10, CIFAR-100, Tiny ImageNet, and ImageNet-100 datasets, when against projected gradient descent PGD10 attack with perturbation budget 8/255. Furthermore, our proposed method also achieves state-of-the-art robust accuracy against other attacks. Code is available at https://github.com/bookman233/TDAT.
Large-scale pre-training has proven to be an effective method for improving performance across different tasks. Current person search methods use ImageNet pre-trained models for feature extraction, yet it is not an optimal solution due to the gap between the pre-training task and person search task (as a downstream task). Therefore, in this paper, we focus on pre-training for person search, which involves detecting and re-identifying individuals simultaneously. Although labeled data for person search is scarce, datasets for two sub-tasks person detection and re-identification are relatively abundant. To this end, we propose a hybrid pre-training framework specifically designed for person search using sub-task data only. It consists of a hybrid learning paradigm that handles data with different kinds of supervisions, and an intra-task alignment module that alleviates domain discrepancy under limited resources. To the best of our knowledge, this is the first work that investigates how to support full-task pre-training using sub-task data. Extensive experiments demonstrate that our pre-trained model can achieve significant improvements across diverse protocols, such as person search method, fine-tuning data, pre-training data and model backbone. For example, our model improves ResNet50 based NAE by 10.3% relative improvement w.r.t. mAP. Our code and pre-trained models are released for plug-and-play usage to the person search community (https://github.com/personsearch/PretrainPS).
A comprehensive understanding of videos is inseparable from describing the action with its contextual action-object interactions. However, many current video understanding tasks prioritize general action classification and overlook the actors and relationships that shape the nature of the action, resulting in a superficial understanding of the action. Motivated by this, we introduce Open-vocabulary Video Relation Extraction (OVRE), a novel task that views action understanding through the lens of action-centric relation triplets. OVRE focuses on pairwise relations that take part in the action and describes these relation triplets with natural languages. Moreover, we curate the Moments-OVRE dataset, which comprises 180K videos with action-centric relation triplets, sourced from a multi-label action classification dataset. With Moments-OVRE, we further propose a cross-modal mapping model to generate relation triplets as a sequence. Finally, we benchmark existing cross-modal generation models on the new task of OVRE. Our code and dataset are available at https://github.com/Iriya99/OVRE.
Towards Efficient and Effective Text-to-Video Retrieval with Coarse-to-Fine Visual Representation Learning
PDF ↗In recent years, text-to-video retrieval methods based on CLIP have experienced rapid development. The primary direction of evolution is to exploit the much wider gamut of visual and textual cues to achieve alignment. Concretely, those methods with impressive performance often design a heavy fusion block for sentence (words)-video (frames) interaction, regardless of the prohibitive computation complexity. Nevertheless, these approaches are not optimal in terms of feature utilization and retrieval efficiency. To address this issue, we adopt multi-granularity visual feature learning, ensuring the model's comprehensiveness in capturing visual content features spanning from abstract to detailed levels during the training phase. To better leverage the multi-granularity features, we devise a two-stage retrieval architecture in the retrieval phase. This solution ingeniously balances the coarse and fine granularity of retrieval content. Moreover, it also strikes a harmonious equilibrium between retrieval effectiveness and efficiency. Specifically, in training phase, we design a parameter-free text-gated interaction block (TIB) for fine-grained video representation learning and embed an extra Pearson Constraint to optimize cross-modal representation learning. In retrieval phase, we use coarse-grained video representations for fast recall of top-k candidates, which are then reranked by fine-grained video representations. Extensive experiments on four benchmarks demonstrate the efficiency and effectiveness. Notably, our method achieves comparable performance with the current state-of-the-art methods while being nearly 50 times faster.
Image composition refers to inserting a foreground object into a background image to obtain a composite image. In this work, we focus on generating plausible shadows for the inserted foreground object to make the composite image more realistic. To supplement the existing small-scale dataset, we create a large-scale dataset called RdSOBA with rendering techniques. Moreover, we design a two-stage network named DMASNet with decomposed mask prediction and attentive shadow filling. Specifically, in the first stage, we decompose shadow mask prediction into box prediction and shape prediction. In the second stage, we attend to reference background shadow pixels to fill the foreground shadow. Abundant experiments prove that our DMASNet achieves better visual effects and generalizes well to real composite images.
Generative models (GMs) have received increasing research interest for their remarkable capacity to achieve comprehensive understanding. However, their potential application in the domain of multi-modal tracking has remained unexplored. In this context, we seek to uncover the potential of harnessing generative techniques to address the critical challenge, information fusion, in multi-modal tracking. In this paper, we delve into two prominent GM techniques, namely, Conditional Generative Adversarial Networks (CGANs) and Diffusion Models (DMs). Different from the standard fusion process where the features from each modality are directly fed into the fusion block, we combine these multi-modal features with random noise in the GM framework, effectively transforming the original training samples into harder instances. This design excels at extracting discriminative clues from the features, enhancing the ultimate tracking performance. Based on this, we conduct extensive experiments across two multi-modal tracking tasks, three baseline methods, and four challenging benchmarks. The experimental results demonstrate that the proposed generative-based fusion mechanism achieves state-of-the-art performance by setting new records on GTOT, LasHeR and RGBD1K. Code will be available at https://github.com/Zhangyong-Tang/GMMT.
Context-I2W: Mapping Images to Context-Dependent Words for Accurate Zero-Shot Composed Image Retrieval
PDF ↗Different from the Composed Image Retrieval task that requires expensive labels for training task-specific models, Zero-Shot Composed Image Retrieval (ZS-CIR) involves diverse tasks with a broad range of visual content manipulation intent that could be related to domain, scene, object, and attribute. The key challenge for ZS-CIR tasks is to learn a more accurate image representation that has adaptive attention to the reference image for various manipulation descriptions. In this paper, we propose a novel context-dependent mapping network, named Context-I2W, for adaptively converting description-relevant Image information into a pseudo-word token composed of the description for accurate ZS-CIR. Specifically, an Intent View Selector first dynamically learns a rotation rule to map the identical image to a task-specific manipulation view. Then a Visual Target Extractor further captures local information covering the main targets in ZS-CIR tasks under the guidance of multiple learnable queries. The two complementary modules work together to map an image to a context-dependent pseudo-word token without extra supervision. Our model shows strong generalization ability on four ZS-CIR tasks, including domain conversion, object composition, object manipulation, and attribute manipulation. It obtains consistent and significant performance boosts ranging from 1.88% to 3.60% over the best methods and achieves new state-of-the-art results on ZS-CIR. Our code is available at https://anonymous.4open.science/r/Context-I2W-4224/.
The popularity of pre-trained large models has revolutionized downstream tasks across diverse fields, such as language, vision, and multi-modality. To minimize the adaption cost for downstream tasks, many Parameter-Efficient Fine-Tuning (PEFT) techniques are proposed for language and 2D image pre-trained models. However, the specialized PEFT method for 3D pre-trained models is still under-explored. To this end, we introduce Point-PEFT, a novel framework for adapting point cloud pre-trained models with minimal learnable parameters. Specifically, for a pre-trained 3D model, we freeze most of its parameters, and only tune the newly added PEFT modules on downstream tasks, which consist of a Point-prior Prompt and a Geometry-aware Adapter. The Point-prior Prompt adopts a set of learnable prompt tokens, for which we propose to construct a memory bank with domain-specific knowledge, and utilize a parameter-free attention to enhance the prompt tokens. The Geometry-aware Adapter aims to aggregate point cloud features within spatial neighborhoods to capture fine-grained geometric information through local interactions. Extensive experiments indicate that our Point-PEFT can achieve better performance than the full fine-tuning on various downstream tasks, while using only 5% of the trainable parameters, demonstrating the efficiency and effectiveness of our approach. Code is released at https://github.com/Ivan-Tang-3D/Point-PEFT.
Recent research understands the residual networks from a new perspective of the implicit ensemble model. From this view, previous methods such as stochastic depth and stimulative training have further improved the performance of the residual network by sampling and training of its subnets. However, they both use the same supervision for all subnets of different capacities and neglect the valuable knowledge generated by subnets during training. In this manuscript, we mitigate the significant knowledge distillation gap caused by using the same kind of supervision and advocate leveraging the subnets to provide diverse knowledge. Based on this motivation, we propose a group knowledge based training framework for boosting the performance of residual networks. Specifically, we implicitly divide all subnets into hierarchical groups by subnet-in-subnet sampling, aggregate the knowledge of different subnets in each group during training, and exploit upper-level group knowledge to supervise lower-level subnet group. Meanwhile, we also develop a subnet sampling strategy that naturally samples larger subnets, which are found to be more helpful than smaller subnets in boosting performance for hierarchical groups. Compared with typical subnet training and other methods, our method achieves the best efficiency and performance trade-offs on multiple datasets and network structures. The code is at https://github.com/tsj-001/AAAI24-GKT.
As a critical clue of video super-resolution (VSR), inter-frame alignment significantly impacts overall performance. However, accurate pixel-level alignment is a challenging task due to the intricate motion interweaving in the video. In response to this issue, we introduce a novel paradigm for VSR named Semantic Lens, predicated on semantic priors drawn from degraded videos. Specifically, video is modeled as instances, events, and scenes via a Semantic Extractor. Those semantics assist the Pixel Enhancer in understanding the recovered contents and generating more realistic visual results. The distilled global semantics embody the scene information of each frame, while the instance-specific semantics assemble the spatial-temporal contexts related to each instance. Furthermore, we devise a Semantics-Powered Attention Cross-Embedding (SPACE) block to bridge the pixel-level features with semantic knowledge, composed of a Global Perspective Shifter (GPS) and an Instance-Specific Semantic Embedding Encoder (ISEE). Concretely, the GPS module generates pairs of affine transformation parameters for pixel-level feature modulation conditioned on global semantics. After that the ISEE module harnesses the attention mechanism to align the adjacent frames in the instance-centric semantic space. In addition, we incorporate a simple yet effective pre-alignment module to alleviate the difficulty of model training. Extensive experiments demonstrate the superiority of our model over existing state-of-the-art VSR methods.
Defocus blur, due to spatially-varying sizes and shapes, is hard to remove. Existing methods either are unable to effectively handle irregular defocus blur or fail to generalize well on other datasets. In this work, we propose a divide-and-conquer approach to tackling this issue, which gives rise to a novel end-to-end deep learning method, called prior-and-prediction inverse kernel transformer (P2IKT), for single image defocus deblurring. Since most defocus blur can be approximated as Gaussian blur or its variants, we construct an inverse Gaussian kernel module in our method to enhance its generalization ability. At the same time, an inverse kernel prediction module is introduced in order to flexibly address the irregular blur that cannot be approximated by Gaussian blur. We further design a scale recurrent transformer, which estimates mixing coefficients for adaptively combining the results from the two modules and runs the scale recurrent ``coarse-to-fine" procedure for progressive defocus deblurring. Extensive experimental results demonstrate that our P2IKT outperforms previous methods in terms of PSNR on multiple defocus deblurring datasets.
Once and for All: Universal Transferable Adversarial Perturbation against Deep Hashing-Based Facial Image Retrieval
PDF ↗Deep Hashing (DH)-based image retrieval has been widely applied to face-matching systems due to its accuracy and efficiency. However, this convenience comes with an increased risk of privacy leakage. DH models inherit the vulnerability to adversarial attacks, which can be used to prevent the retrieval of private images. Existing adversarial attacks against DH typically target a single image or a specific class of images, lacking universal adversarial perturbation for the entire hash dataset. In this paper, we propose the first universal transferable adversarial perturbation against DH-based facial image retrieval, a single perturbation can protect all images. Specifically, we explore the relationship between clusters learned by different DH models and define the optimization objective of universal perturbation as leaving from the overall hash center. To mitigate the challenge of single-objective optimization, we randomly obtain sub-cluster centers and further propose sub-task-based meta-learning to aid in overall optimization. We test our method with popular facial datasets and DH models, indicating impressive cross-image, -identity, -model, and -scheme universal anti-retrieval performance. Compared to state-of-the-art methods, our performance is competitive in white-box settings and exhibits significant improvements of 10%-70% in transferability in all black-box settings.
Adversarial attacks on 3D point clouds often exhibit unsatisfactory imperceptibility, which primarily stems from the disregard for manifold-aware distortion, i.e., distortion of the underlying 2-manifold surfaces. In this paper, we develop novel manifold constraints to reduce such distortion, aiming to enhance the imperceptibility of adversarial attacks on 3D point clouds. Specifically, we construct a bijective manifold mapping between point clouds and a simple parameter shape using an invertible auto-encoder. Consequently, manifold-aware distortion during attacks can be captured within the parameter space. By enforcing manifold constraints that preserve local properties of the parameter shape, manifold-aware distortion is effectively mitigated, ultimately leading to enhanced imperceptibility. Extensive experiments demonstrate that integrating manifold constraints into conventional adversarial attack solutions yields superior imperceptibility, outperforming the state-of-the-art methods.
Video compression relies heavily on exploiting the temporal redundancy between video frames, which is usually achieved by estimating and using the motion information. The motion information is represented as optical flows in most of the existing deep video compression networks. Indeed, these networks often adopt pre-trained optical flow estimation networks for motion estimation. The optical flows, however, may be less suitable for video compression due to the following two factors. First, the optical flow estimation networks were trained to perform inter-frame prediction as accurately as possible, but the optical flows themselves may cost too many bits to encode. Second, the optical flow estimation networks were trained on synthetic data, and may not generalize well enough to real-world videos. We address the twofold limitations by enhancing the optical flows in two stages: offline and online. In the offline stage, we fine-tune a trained optical flow estimation network with the motion information provided by a traditional (non-deep) video compression scheme, e.g. H.266/VVC, as we believe the motion information of H.266/VVC achieves a better rate-distortion trade-off. In the online stage, we further optimize the latent features of the optical flows with a gradient descent-based algorithm for the video to be compressed, so as to enhance the adaptivity of the optical flows. We conduct experiments on two state-of-the-art deep video compression schemes, DCVC and DCVC-DC. Experimental results demonstrate that the proposed offline and online enhancement together achieves on average 13.4% bitrate saving for DCVC and 4.1% bitrate saving for DCVC-DC on the tested videos, without increasing the model or computational complexity of the decoder side.
Deep learning models have the ability to extract rich knowledge from large-scale datasets. However, the sharing of data has become increasingly challenging due to concerns regarding data copyright and privacy. Consequently, this hampers the effective transfer of knowledge from existing data to novel downstream tasks and concepts. Zero-shot learning (ZSL) approaches aim to recognize new classes by transferring semantic knowledge learned from base classes. However, traditional generative ZSL methods often require access to real images from base classes and rely on manually annotated attributes, which presents challenges in terms of data restrictions and model scalability. To this end, this paper tackles a challenging and practical problem dubbed as data-free zero-shot learning (DFZSL), where only the CLIP-based base classes data pre-trained classifier is available for zero-shot classification. Specifically, we propose a generic framework for DFZSL, which consists of three main components. Firstly, to recover the virtual features of the base data, we model the CLIP features of base class images as samples from a von Mises-Fisher (vMF) distribution based on the pre-trained classifier. Secondly, we leverage the text features of CLIP as low-cost semantic information and propose a feature-language prompt tuning (FLPT) method to further align the virtual image features and textual features. Thirdly, we train a conditional generative model using the well-aligned virtual image features and corresponding semantic text features, enabling the generation of new classes features and achieve better zero-shot generalization. Our framework has been evaluated on five commonly used benchmarks for generalized ZSL, as well as 11 benchmarks for the base-to-new ZSL. The results demonstrate the superiority and effectiveness of our approach. Our code is available in https://github.com/ylong4/DFZSL.
Data augmentation has been recently leveraged as an effective regularizer in various vision-language deep neural networks. However, in text-to-image synthesis (T2Isyn), current augmentation wisdom still suffers from the semantic mismatch between augmented paired data. Even worse, semantic collapse may occur when generated images are less semantically constrained. In this paper, we develop a novel Semantic-aware Data Augmentation (SADA) framework dedicated to T2Isyn. In particular, we propose to augment texts in the semantic space via an Implicit Textual Semantic Preserving Augmentation, in conjunction with a specifically designed Image Semantic Regularization Loss as Generated Image Semantic Conservation, to cope well with semantic mismatch and collapse. As one major contribution, we theoretically show that Implicit Textual Semantic Preserving Augmentation can certify better text-image consistency while Image Semantic Regularization Loss regularizing the semantics of generated images would avoid semantic collapse and enhance image quality. Extensive experiments validate that SADA enhances text-image consistency and improves image quality significantly in T2Isyn models across various backbones. Especially, incorporating SADA during the tuning process of Stable Diffusion models also yields performance improvements.