Zheng-Jun Zha

Shaofei Cai, Liang Li, Xinzhe Han et al. · pku

Graph neural architecture search has sparked much attention as Graph Neural Networks (GNNs) have shown powerful reasoning capability in many relational tasks. However, the currently used graph search space overemphasizes learning node features and neglects mining hierarchical relational information. Moreover, due to diverse mechanisms in the message passing, the graph search space is much larger than that of CNNs. This hinders the straightforward application of classical search strategies for exploring complicated graph search space. We propose Automatic Relation-aware Graph Network Proliferation (ARGNP) for efficiently searching GNNs with a relation-guided message passing mechanism. Specifically, we first devise a novel dual relation-aware graph search space that comprises both node and relation learning operations. These operations can extract hierarchical node/relational information and provide anisotropic guidance for message passing on a graph. Second, analogous to cell proliferation, we design a network proliferation search paradigm to progressively determine the GNN architectures by iteratively performing network division and differentiation. The experiments on six datasets for four graph learning tasks demonstrate that GNNs produced by our method are superior to the current state-of-the-art hand-crafted and search-based GNNs. Codes are available at https://github.com/phython96/ARGNP.

Yuhang Yang, Wei Zhai, Hongchen Luo et al.

Grounding 3D object affordance seeks to locate objects' ''action possibilities'' regions in the 3D space, which serves as a link between perception and operation for embodied agents. Existing studies primarily focus on connecting visual affordances with geometry structures, e.g. relying on annotations to declare interactive regions of interest on the object and establishing a mapping between the regions and affordances. However, the essence of learning object affordance is to understand how to use it, and the manner that detaches interactions is limited in generalization. Normally, humans possess the ability to perceive object affordances in the physical world through demonstration images or videos. Motivated by this, we introduce a novel task setting: grounding 3D object affordance from 2D interactions in images, which faces the challenge of anticipating affordance through interactions of different sources. To address this problem, we devise a novel Interaction-driven 3D Affordance Grounding Network (IAG), which aligns the region feature of objects from different sources and models the interactive contexts for 3D object affordance grounding. Besides, we collect a Point-Image Affordance Dataset (PIAD) to support the proposed task. Comprehensive experiments on PIAD demonstrate the reliability of the proposed task and the superiority of our method. The project is available at https://github.com/yyvhang/IAGNet.

Pingyu Wu, Wei Zhai, Yang Cao et al.

Weakly supervised object localization (WSOL) is a challenging task aiming to localize objects with only image-level supervision. Recent works apply visual transformer to WSOL and achieve significant success by exploiting the long-range feature dependency in self-attention mechanism. However, existing transformer-based methods synthesize the classification feature maps as the localization map, which leads to optimization conflicts between classification and localization tasks. To address this problem, we propose to learn a task-specific spatial-aware token (SAT) to condition localization in a weakly supervised manner. Specifically, a spatial token is first introduced in the input space to aggregate representations for localization task. Then a spatial aware attention module is constructed, which allows spatial token to generate foreground probabilities of different patches by querying and to extract localization knowledge from the classification task. Besides, for the problem of sparse and unbalanced pixel-level supervision obtained from the image-level label, two spatial constraints, including batch area loss and normalization loss, are designed to compensate and enhance this supervision. Experiments show that the proposed SAT achieves state-of-the-art performance on both CUB-200 and ImageNet, with 98.45% and 73.13% GT-known Loc, respectively. Even under the extreme setting of using only 1 image per class from ImageNet for training, SAT already exceeds the SOTA method by 2.1% GT-known Loc. Code and models are available at https://github.com/wpy1999/SAT.

Yucheng Zhao, Chong Luo, Chuanxin Tang et al.

Video understanding tasks have traditionally been modeled by two separate architectures, specially tailored for two distinct tasks. Sequence-based video tasks, such as action recognition, use a video backbone to directly extract spatiotemporal features, while frame-based video tasks, such as multiple object tracking (MOT), rely on single fixed-image backbone to extract spatial features. In contrast, we propose to unify video understanding tasks into one novel streaming video architecture, referred to as Streaming Vision Transformer (S-ViT). S-ViT first produces frame-level features with a memory-enabled temporally-aware spatial encoder to serve the frame-based video tasks. Then the frame features are input into a task-related temporal decoder to obtain spatiotemporal features for sequence-based tasks. The efficiency and efficacy of S-ViT is demonstrated by the state-of-the-art accuracy in the sequence-based action recognition task and the competitive advantage over conventional architecture in the frame-based MOT task. We believe that the concept of streaming video model and the implementation of S-ViT are solid steps towards a unified deep learning architecture for video understanding. Code will be available at https://github.com/yuzhms/Streaming-Video-Model.

Yangyang Li, Wei Zhai, Yang Cao et al.

Camouflage is a common visual phenomenon, which refers to hiding the foreground objects into the background images, making them briefly invisible to the human eye. Previous work has typically been implemented by an iterative optimization process. However, these methods struggle in 1) efficiently generating camouflage images using foreground and background with arbitrary structure; 2) camouflaging foreground objects to regions with multiple appearances (e.g. the junction of the vegetation and the mountains), which limit their practical application. To address these problems, this paper proposes a novel Location-free Camouflage Generation Network (LCG-Net) that fuse high-level features of foreground and background image, and generate result by one inference. Specifically, a Position-aligned Structure Fusion (PSF) module is devised to guide structure feature fusion based on the point-to-point structure similarity of foreground and background, and introduce local appearance features point-by-point. To retain the necessary identifiable features, a new immerse loss is adopted under our pipeline, while a background patch appearance loss is utilized to ensure that the hidden objects look continuous and natural at regions with multiple appearances. Experiments show that our method has results as satisfactory as state-of-the-art in the single-appearance regions and are less likely to be completely invisible, but far exceed the quality of the state-of-the-art in the multi-appearance regions. Moreover, our method is hundreds of times faster than previous methods. Benefitting from the unique advantages of our method, we provide some downstream applications for camouflage generation, which show its potential. The related code and dataset will be released at https://github.com/Tale17/LCG-Net.

Yunbin Tu, Liang Li, Li Su et al.

Change captioning aims to describe the difference between a pair of similar images. Its key challenge is how to learn a stable difference representation under pseudo changes caused by viewpoint change. In this paper, we address this by proposing a self-supervised cross-view representation reconstruction (SCORER) network. Concretely, we first design a multi-head token-wise matching to model relationships between cross-view features from similar/dissimilar images. Then, by maximizing cross-view contrastive alignment of two similar images, SCORER learns two view-invariant image representations in a self-supervised way. Based on these, we reconstruct the representations of unchanged objects by cross-attention, thus learning a stable difference representation for caption generation. Further, we devise a cross-modal backward reasoning to improve the quality of caption. This module reversely models a ``hallucination'' representation with the caption and ``before'' representation. By pushing it closer to the ``after'' representation, we enforce the caption to be informative about the difference in a self-supervised manner. Extensive experiments show our method achieves the state-of-the-art results on four datasets. The code is available at https://github.com/tuyunbin/SCORER.

Kai Zhu, Wei Zhai, Yang Cao et al.

Non-exemplar class-incremental learning is to recognize both the old and new classes when old class samples cannot be saved. It is a challenging task since representation optimization and feature retention can only be achieved under supervision from new classes. To address this problem, we propose a novel self-sustaining representation expansion scheme. Our scheme consists of a structure reorganization strategy that fuses main-branch expansion and side-branch updating to maintain the old features, and a main-branch distillation scheme to transfer the invariant knowledge. Furthermore, a prototype selection mechanism is proposed to enhance the discrimination between the old and new classes by selectively incorporating new samples into the distillation process. Extensive experiments on three benchmarks demonstrate significant incremental performance, outperforming the state-of-the-art methods by a margin of 3%, 3% and 6%, respectively.

Wei Zhai, Pingyu Wu, Kai Zhu et al.

Weakly supervised object localization and semantic segmentation aim to localize objects using only image-level labels. Recently, a new paradigm has emerged by generating a foreground prediction map (FPM) to achieve pixel-level localization. While existing FPM-based methods use cross-entropy to evaluate the foreground prediction map and to guide the learning of the generator, this paper presents two astonishing experimental observations on the object localization learning process: For a trained network, as the foreground mask expands, 1) the cross-entropy converges to zero when the foreground mask covers only part of the object region. 2) The activation value continuously increases until the foreground mask expands to the object boundary. Therefore, to achieve a more effective localization performance, we argue for the usage of activation value to learn more object regions. In this paper, we propose a Background Activation Suppression (BAS) method. Specifically, an Activation Map Constraint (AMC) module is designed to facilitate the learning of generator by suppressing the background activation value. Meanwhile, by using foreground region guidance and area constraint, BAS can learn the whole region of the object. In the inference phase, we consider the prediction maps of different categories together to obtain the final localization results. Extensive experiments show that BAS achieves significant and consistent improvement over the baseline methods on the CUB-200-2011 and ILSVRC datasets. In addition, our method also achieves state-of-the-art weakly supervised semantic segmentation performance on the PASCAL VOC 2012 and MS COCO 2014 datasets. Code and models are available at https://github.com/wpy1999/BAS-Extension.

Zhenhuan Liu, Liang Li, Huajie Jiang et al. · nvidia

In recent years, creative content generations like style transfer and neural photo editing have attracted more and more attention. Among these, cartoonization of real-world scenes has promising applications in entertainment and industry. Different from image translations focusing on improving the style effect of generated images, video cartoonization has additional requirements on the temporal consistency. In this paper, we propose a spatially-adaptive semantic alignment framework with perceptual motion consistency for coherent video cartoonization in an unsupervised manner. The semantic alignment module is designed to restore deformation of semantic structure caused by spatial information lost in the encoder-decoder architecture. Furthermore, we devise the spatio-temporal correlative map as a style-independent, global-aware regularization on the perceptual motion consistency. Deriving from similarity measurement of high-level features in photo and cartoon frames, it captures global semantic information beyond raw pixel-value in optical flow. Besides, the similarity measurement disentangles temporal relationships from domain-specific style properties, which helps regularize the temporal consistency without hurting style effects of cartoon images. Qualitative and quantitative experiments demonstrate our method is able to generate highly stylistic and temporal consistent cartoon videos.

Jiayu Xiao, Liang Li, Chaofei Wang et al.

Training a generative adversarial network (GAN) with limited data has been a challenging task. A feasible solution is to start with a GAN well-trained on a large scale source domain and adapt it to the target domain with a few samples, termed as few shot generative model adaption. However, existing methods are prone to model overfitting and collapse in extremely few shot setting (less than 10). To solve this problem, we propose a relaxed spatial structural alignment method to calibrate the target generative models during the adaption. We design a cross-domain spatial structural consistency loss comprising the self-correlation and disturbance correlation consistency loss. It helps align the spatial structural information between the synthesis image pairs of the source and target domains. To relax the cross-domain alignment, we compress the original latent space of generative models to a subspace. Image pairs generated from the subspace are pulled closer. Qualitative and quantitative experiments show that our method consistently surpasses the state-of-the-art methods in few shot setting.

Zhipeng Huang, Jiawei Liu, Liang Li et al.

RGB-infrared person re-identification is an emerging cross-modality re-identification task, which is very challenging due to significant modality discrepancy between RGB and infrared images. In this work, we propose a novel modality-adaptive mixup and invariant decomposition (MID) approach for RGB-infrared person re-identification towards learning modality-invariant and discriminative representations. MID designs a modality-adaptive mixup scheme to generate suitable mixed modality images between RGB and infrared images for mitigating the inherent modality discrepancy at the pixel-level. It formulates modality mixup procedure as Markov decision process, where an actor-critic agent learns dynamical and local linear interpolation policy between different regions of cross-modality images under a deep reinforcement learning framework. Such policy guarantees modality-invariance in a more continuous latent space and avoids manifold intrusion by the corrupted mixed modality samples. Moreover, to further counter modality discrepancy and enforce invariant visual semantics at the feature-level, MID employs modality-adaptive convolution decomposition to disassemble a regular convolution layer into modality-specific basis layers and a modality-shared coefficient layer. Extensive experimental results on two challenging benchmarks demonstrate superior performance of MID over state-of-the-art methods.

Zhipeng Huang, Zhizheng Zhang, Cuiling Lan et al.

Recent vision transformers, large-kernel CNNs and MLPs have attained remarkable successes in broad vision tasks thanks to their effective information fusion in the global scope. However, their efficient deployments, especially on mobile devices, still suffer from noteworthy challenges due to the heavy computational costs of self-attention mechanisms, large kernels, or fully connected layers. In this work, we apply conventional convolution theorem to deep learning for addressing this and reveal that adaptive frequency filters can serve as efficient global token mixers. With this insight, we propose Adaptive Frequency Filtering (AFF) token mixer. This neural operator transfers a latent representation to the frequency domain via a Fourier transform and performs semantic-adaptive frequency filtering via an elementwise multiplication, which mathematically equals to a token mixing operation in the original latent space with a dynamic convolution kernel as large as the spatial resolution of this latent representation. We take AFF token mixers as primary neural operators to build a lightweight neural network, dubbed AFFNet. Extensive experiments demonstrate the effectiveness of our proposed AFF token mixer and show that AFFNet achieve superior accuracy and efficiency trade-offs compared to other lightweight network designs on broad visual tasks, including visual recognition and dense prediction tasks.

Xuejing Liu, Liang Li, Shuhui Wang et al.

Weakly supervised Referring Expression Grounding (REG) aims to ground a particular target in an image described by a language expression while lacking the correspondence between target and expression. Two main problems exist in weakly supervised REG. First, the lack of region-level annotations introduces ambiguities between proposals and queries. Second, most previous weakly supervised REG methods ignore the discriminative location and context of the referent, causing difficulties in distinguishing the target from other same-category objects. To address the above challenges, we design an entity-enhanced adaptive reconstruction network (EARN). Specifically, EARN includes three modules: entity enhancement, adaptive grounding, and collaborative reconstruction. In entity enhancement, we calculate semantic similarity as supervision to select the candidate proposals. Adaptive grounding calculates the ranking score of candidate proposals upon subject, location and context with hierarchical attention. Collaborative reconstruction measures the ranking result from three perspectives: adaptive reconstruction, language reconstruction and attribute classification. The adaptive mechanism helps to alleviate the variance of different referring expressions. Experiments on five datasets show EARN outperforms existing state-of-the-art methods. Qualitative results demonstrate that the proposed EARN can better handle the situation where multiple objects of a particular category are situated together.

Ruili Feng, Kecheng Zheng, Yukun Huang et al.

The rank of neural networks measures information flowing across layers. It is an instance of a key structural condition that applies across broad domains of machine learning. In particular, the assumption of low-rank feature representations leads to algorithmic developments in many architectures. For neural networks, however, the intrinsic mechanism that yields low-rank structures remains vague and unclear. To fill this gap, we perform a rigorous study on the behavior of network rank, focusing particularly on the notion of rank deficiency. We theoretically establish a universal monotonic decreasing property of network rank from the basic rules of differential and algebraic composition, and uncover rank deficiency of network blocks and deep function coupling. By virtue of our numerical tools, we provide the first empirical analysis of the per-layer behavior of network rank in practical settings, i.e., ResNets, deep MLPs, and Transformers on ImageNet. These empirical results are in direct accord with our theory. Furthermore, we reveal a novel phenomenon of independence deficit caused by the rank deficiency of deep networks, where classification confidence of a given category can be linearly decided by the confidence of a handful of other categories. The theoretical results of this work, together with the empirical findings, may advance understanding of the inherent principles of deep neural networks.

Gaoxiang Cong, Liang Li, Yuankai Qi et al.

Given a piece of text, a video clip and a reference audio, the movie dubbing (also known as visual voice clone V2C) task aims to generate speeches that match the speaker's emotion presented in the video using the desired speaker voice as reference. V2C is more challenging than conventional text-to-speech tasks as it additionally requires the generated speech to exactly match the varying emotions and speaking speed presented in the video. Unlike previous works, we propose a novel movie dubbing architecture to tackle these problems via hierarchical prosody modelling, which bridges the visual information to corresponding speech prosody from three aspects: lip, face, and scene. Specifically, we align lip movement to the speech duration, and convey facial expression to speech energy and pitch via attention mechanism based on valence and arousal representations inspired by recent psychology findings. Moreover, we design an emotion booster to capture the atmosphere from global video scenes. All these embeddings together are used to generate mel-spectrogram and then convert to speech waves via existing vocoder. Extensive experimental results on the Chem and V2C benchmark datasets demonstrate the favorable performance of the proposed method. The source code and trained models will be released to the public.

Jiawei Liu, Zhipeng Huang, Liang Li et al.

Generalizable person re-identification aims to learn a model with only several labeled source domains that can perform well on unseen domains. Without access to the unseen domain, the feature statistics of the batch normalization (BN) layer learned from a limited number of source domains is doubtlessly biased for unseen domain. This would mislead the feature representation learning for unseen domain and deteriorate the generalizaiton ability of the model. In this paper, we propose a novel Debiased Batch Normalization via Gaussian Process approach (GDNorm) for generalizable person re-identification, which models the feature statistic estimation from BN layers as a dynamically self-refining Gaussian process to alleviate the bias to unseen domain for improving the generalization. Specifically, we establish a lightweight model with multiple set of domain-specific BN layers to capture the discriminability of individual source domain, and learn the corresponding parameters of the domain-specific BN layers. These parameters of different source domains are employed to deduce a Gaussian process. We randomly sample several paths from this Gaussian process served as the BN estimations of potential new domains outside of existing source domains, which can further optimize these learned parameters from source domains, and estimate more accurate Gaussian process by them in return, tending to real data distribution. Even without a large number of source domains, GDNorm can still provide debiased BN estimation by using the mean path of the Gaussian process, while maintaining low computational cost during testing. Extensive experiments demonstrate that our GDNorm effectively improves the generalization ability of the model on unseen domain.

Naishan Zheng, Jie Huang, Qi Zhu et al.

Low-light image enhancement is an inherently subjective process whose targets vary with the user's aesthetic. Motivated by this, several personalized enhancement methods have been investigated. However, the enhancement process based on user preferences in these techniques is invisible, i.e., a "black box". In this work, we propose an intelligible unsupervised personalized enhancer (iUPEnhancer) for low-light images, which establishes the correlations between the low-light and the unpaired reference images with regard to three user-friendly attributions (brightness, chromaticity, and noise). The proposed iUP-Enhancer is trained with the guidance of these correlations and the corresponding unsupervised loss functions. Rather than a "black box" process, our iUP-Enhancer presents an intelligible enhancement process with the above attributions. Extensive experiments demonstrate that the proposed algorithm produces competitive qualitative and quantitative results while maintaining excellent flexibility and scalability. This can be validated by personalization with single/multiple references, cross-attribution references, or merely adjusting parameters.

Jinze Chen, Yang Wang, Yang Cao et al.

Dynamic Vision Sensor (DVS) can asynchronously output the events reflecting apparent motion of objects with microsecond resolution, and shows great application potential in monitoring and other fields. However, the output event stream of existing DVS inevitably contains background activity noise (BA noise) due to dark current and junction leakage current, which will affect the temporal correlation of objects, resulting in deteriorated motion estimation performance. Particularly, the existing filter-based denoising methods cannot be directly applied to suppress the noise in event stream, since there is no spatial correlation. To address this issue, this paper presents a novel progressive framework, in which a Motion Estimation (ME) module and an Event Denoising (ED) module are jointly optimized in a mutually reinforced manner. Specifically, based on the maximum sharpness criterion, ME module divides the input event into several segments by adaptive clustering in a motion compensating warp field, and captures the temporal correlation of event stream according to the clustered motion parameters. Taking temporal correlation as guidance, ED module calculates the confidence that each event belongs to real activity events, and transmits it to ME module to update energy function of motion segmentation for noise suppression. The two steps are iteratively updated until stable motion segmentation results are obtained. Extensive experimental results on both synthetic and real datasets demonstrate the superiority of our proposed approaches against the State-Of-The-Art (SOTA) methods.

Kecheng Zheng, Wei Wu, Ruili Feng et al.

Prompt tuning and adapter tuning have shown great potential in transferring pre-trained vision-language models (VLMs) to various downstream tasks. In this work, we design a new type of tuning method, termed as regularized mask tuning, which masks the network parameters through a learnable selection. Inspired by neural pathways, we argue that the knowledge required by a downstream task already exists in the pre-trained weights but just gets concealed in the upstream pre-training stage. To bring the useful knowledge back into light, we first identify a set of parameters that are important to a given downstream task, then attach a binary mask to each parameter, and finally optimize these masks on the downstream data with the parameters frozen. When updating the mask, we introduce a novel gradient dropout strategy to regularize the parameter selection, in order to prevent the model from forgetting old knowledge and overfitting the downstream data. Experimental results on 11 datasets demonstrate the consistent superiority of our method over previous alternatives. It is noteworthy that we manage to deliver 18.73% performance improvement compared to the zero-shot CLIP via masking an average of only 2.56% parameters. Furthermore, our method is synergistic with most existing parameter-efficient tuning methods and can boost the performance on top of them. Project page can be found here (https://wuw2019.github.io/R-AMT/).

Yurui Zhu, Xueyang Fu, Peng-Tao Jiang et al.

This research focuses on the issue of single-image reflection removal (SIRR) in real-world conditions, examining it from two angles: the collection pipeline of real reflection pairs and the perception of real reflection locations. We devise an advanced reflection collection pipeline that is highly adaptable to a wide range of real-world reflection scenarios and incurs reduced costs in collecting large-scale aligned reflection pairs. In the process, we develop a large-scale, high-quality reflection dataset named Reflection Removal in the Wild (RRW). RRW contains over 14,950 high-resolution real-world reflection pairs, a dataset forty-five times larger than its predecessors. Regarding perception of reflection locations, we identify that numerous virtual reflection objects visible in reflection images are not present in the corresponding ground-truth images. This observation, drawn from the aligned pairs, leads us to conceive the Maximum Reflection Filter (MaxRF). The MaxRF could accurately and explicitly characterize reflection locations from pairs of images. Building upon this, we design a reflection location-aware cascaded framework, specifically tailored for SIRR. Powered by these innovative techniques, our solution achieves superior performance than current leading methods across multiple real-world benchmarks. Codes and datasets will be publicly available.

Xihao Chen, Zhiwei Xiong, Zhen Cheng et al.

In this paper, we study the problem of stereo matching from a pair of images with different resolutions, e.g., those acquired with a tele-wide camera system. Due to the difficulty of obtaining ground-truth disparity labels in diverse real-world systems, we start from an unsupervised learning perspective. However, resolution asymmetry caused by unknown degradations between two views hinders the effectiveness of the generally assumed photometric consistency. To overcome this challenge, we propose to impose the consistency between two views in a feature space instead of the image space, named feature-metric consistency. Interestingly, we find that, although a stereo matching network trained with the photometric loss is not optimal, its feature extractor can produce degradation-agnostic and matching-specific features. These features can then be utilized to formulate a feature-metric loss to avoid the photometric inconsistency. Moreover, we introduce a self-boosting strategy to optimize the feature extractor progressively, which further strengthens the feature-metric consistency. Experiments on both simulated datasets with various degradations and a self-collected real-world dataset validate the superior performance of the proposed method over existing solutions.

Kecheng Zheng, Yang Cao, Kai Zhu et al.

MLP-like models built entirely upon multi-layer perceptrons have recently been revisited, exhibiting the comparable performance with transformers. It is one of most promising architectures due to the excellent trade-off between network capability and efficiency in the large-scale recognition tasks. However, its generalization performance to heterogeneous tasks is inferior to other architectures (e.g., CNNs and transformers) due to the extensive retention of domain information. To address this problem, we propose a novel frequency-aware MLP architecture, in which the domain-specific features are filtered out in the transformed frequency domain, augmenting the invariant descriptor for label prediction. Specifically, we design an adaptive Fourier filter layer, in which a learnable frequency filter is utilized to adjust the amplitude distribution by optimizing both the real and imaginary parts. A low-rank enhancement module is further proposed to rectify the filtered features by adding the low-frequency components from SVD decomposition. Finally, a momentum update strategy is utilized to stabilize the optimization to fluctuation of model parameters and inputs by the output distillation with weighted historical states. To our best knowledge, we are the first to propose a MLP-like backbone for domain generalization. Extensive experiments on three benchmarks demonstrate significant generalization performance, outperforming the state-of-the-art methods by a margin of 3%, 4% and 9%, respectively.

Ruili Feng, Jie Xiao, Kecheng Zheng et al.

Human can extrapolate well, generalize daily knowledge into unseen scenarios, raise and answer counterfactual questions. To imitate this ability via generative models, previous works have extensively studied explicitly encoding Structural Causal Models (SCMs) into architectures of generator networks. This methodology, however, limits the flexibility of the generator as they must be carefully crafted to follow the causal graph, and demands a ground truth SCM with strong ignorability assumption as prior, which is a nontrivial assumption in many real scenarios. Thus, many current causal GAN methods fail to generate high fidelity counterfactual results as they cannot easily leverage state-of-the-art generative models. In this paper, we propose to study counterfactual synthesis from a new perspective of knowledge extrapolation, where a given knowledge dimension of the data distribution is extrapolated, but the remaining knowledge is kept indistinguishable from the original distribution. We show that an adversarial game with a closed-form discriminator can be used to address the knowledge extrapolation problem, and a novel principal knowledge descent method can efficiently estimate the extrapolated distribution through the adversarial game. Our method enjoys both elegant theoretical guarantees and superior performance in many scenarios.

Jingyi Xie, Jiawei Liu, Zheng-Jun Zha

Fake news spreads at an unprecedented speed, reaches global audiences and poses huge risks to users and communities. Most existing fake news detection algorithms focus on building supervised training models on a large amount of manually labeled data, which is expensive to acquire or often unavailable. In this work, we propose a novel label noise-resistant mean teaching approach (LNMT) for weakly supervised fake news detection. LNMT leverages unlabeled news and feedback comments of users to enlarge the amount of training data and facilitates model training by generating refined labels as weak supervision. Specifically, LNMT automatically assigns initial weak labels to unlabeled samples based on semantic correlation and emotional association between news content and the comments. Moreover, in order to suppress the noises in weak labels, LNMT establishes a mean teacher framework equipped with label propagation and label reliability estimation. The framework measures a weak label similarity matrix between the teacher and student networks, and propagates different valuable weak label information to refine the weak labels. Meanwhile, it exploits the consistency between the output class likelihood vectors of the two networks to evaluate the reliability of the weak labels and incorporates the reliability into model optimization to alleviate the negative effect of noisy weak labels. Extensive experiments show the superior performance of LNMT.

Jiawei Liu, Jingyi Xie, Fanrui Zhang et al.

The explosive growth of rumors with text and images on social media platforms has drawn great attention. Existing studies have made significant contributions to cross-modal information interaction and fusion, but they fail to fully explore hierarchical and complex semantic correlation across different modality content, severely limiting their performance on detecting multi-modal rumor. In this work, we propose a novel knowledge-enhanced hierarchical information correlation learning approach (KhiCL) for multi-modal rumor detection by jointly modeling the basic semantic correlation and high-order knowledge-enhanced entity correlation. Specifically, KhiCL exploits cross-modal joint dictionary to transfer the heterogeneous unimodality features into the common feature space and captures the basic cross-modal semantic consistency and inconsistency by a cross-modal fusion layer. Moreover, considering the description of multi-modal content is narrated around entities, KhiCL extracts visual and textual entities from images and text, and designs a knowledge relevance reasoning strategy to find the shortest semantic relevant path between each pair of entities in external knowledge graph, and absorbs all complementary contextual knowledge of other connected entities in this path for learning knowledge-enhanced entity representations. Furthermore, KhiCL utilizes a signed attention mechanism to model the knowledge-enhanced entity consistency and inconsistency of intra-modality and inter-modality entity pairs by measuring their corresponding semantic relevant distance. Extensive experiments have demonstrated the effectiveness of the proposed method.

Ruili Feng, Kecheng Zheng, Kai Zhu et al.

This work presents two astonishing findings on neural networks learned for large-scale image classification. 1) Given a well-trained model, the logits predicted for some category can be directly obtained by linearly combining the predictions of a few other categories, which we call \textbf{neural dependency}. 2) Neural dependencies exist not only within a single model, but even between two independently learned models, regardless of their architectures. Towards a theoretical analysis of such phenomena, we demonstrate that identifying neural dependencies is equivalent to solving the Covariance Lasso (CovLasso) regression problem proposed in this paper. Through investigating the properties of the problem solution, we confirm that neural dependency is guaranteed by a redundant logit covariance matrix, which condition is easily met given massive categories, and that neural dependency is highly sparse, implying that one category correlates to only a few others. We further empirically show the potential of neural dependencies in understanding internal data correlations, generalizing models to unseen categories, and improving model robustness with a dependency-derived regularizer. Code for this work will be made publicly available.

Xin Di, Long Peng, Peizhe Xia et al.

Burst super-resolution aims to reconstruct high-resolution images with higher quality and richer details by fusing the sub-pixel information from multiple burst low-resolution frames. In BusrtSR, the key challenge lies in extracting the base frame's content complementary sub-pixel details while simultaneously suppressing high-frequency noise disturbance. Existing methods attempt to extract sub-pixels by modeling inter-frame relationships frame by frame while overlooking the mutual correlations among multi-current frames and neglecting the intra-frame interactions, leading to inaccurate and noisy sub-pixels for base frame super-resolution. Further, existing methods mainly employ static upsampling with fixed parameters to improve spatial resolution for all scenes, failing to perceive the sub-pixel distribution difference across multiple frames and cannot balance the fusion weights of different frames, resulting in over-smoothed details and artifacts. To address these limitations, we introduce a novel Query Mamba Burst Super-Resolution (QMambaBSR) network, which incorporates a Query State Space Model (QSSM) and Adaptive Up-sampling module (AdaUp). Specifically, based on the observation that sub-pixels have consistent spatial distribution while random noise is inconsistently distributed, a novel QSSM is proposed to efficiently extract sub-pixels through inter-frame querying and intra-frame scanning while mitigating noise interference in a single step. Moreover, AdaUp is designed to dynamically adjust the upsampling kernel based on the spatial distribution of multi-frame sub-pixel information in the different burst scenes, thereby facilitating the reconstruction of the spatial arrangement of high-resolution details. Extensive experiments on four popular synthetic and real-world benchmarks demonstrate that our method achieves a new state-of-the-art performance.

Cuiyu Liu, Wei Zhai, Yuhang Yang et al.

Grounding 3D scene affordance aims to locate interactive regions in 3D environments, which is crucial for embodied agents to interact intelligently with their surroundings. Most existing approaches achieve this by mapping semantics to 3D instances based on static geometric structure and visual appearance. This passive strategy limits the agent's ability to actively perceive and engage with the environment, making it reliant on predefined semantic instructions. In contrast, humans develop complex interaction skills by observing and imitating how others interact with their surroundings. To empower the model with such abilities, we introduce a novel task: grounding 3D scene affordance from egocentric interactions, where the goal is to identify the corresponding affordance regions in a 3D scene based on an egocentric video of an interaction. This task faces the challenges of spatial complexity and alignment complexity across multiple sources. To address these challenges, we propose the Egocentric Interaction-driven 3D Scene Affordance Grounding (Ego-SAG) framework, which utilizes interaction intent to guide the model in focusing on interaction-relevant sub-regions and aligns affordance features from different sources through a bidirectional query decoder mechanism. Furthermore, we introduce the Egocentric Video-3D Scene Affordance Dataset (VSAD), covering a wide range of common interaction types and diverse 3D environments to support this task. Extensive experiments on VSAD validate both the feasibility of the proposed task and the effectiveness of our approach.

Yukun Huang, Jianan Wang, Ailing Zeng et al.

Leveraging pretrained 2D diffusion models and score distillation sampling (SDS), recent methods have shown promising results for text-to-3D avatar generation. However, generating high-quality 3D avatars capable of expressive animation remains challenging. In this work, we present DreamWaltz-G, a novel learning framework for animatable 3D avatar generation from text. The core of this framework lies in Skeleton-guided Score Distillation and Hybrid 3D Gaussian Avatar representation. Specifically, the proposed skeleton-guided score distillation integrates skeleton controls from 3D human templates into 2D diffusion models, enhancing the consistency of SDS supervision in terms of view and human pose. This facilitates the generation of high-quality avatars, mitigating issues such as multiple faces, extra limbs, and blurring. The proposed hybrid 3D Gaussian avatar representation builds on the efficient 3D Gaussians, combining neural implicit fields and parameterized 3D meshes to enable real-time rendering, stable SDS optimization, and expressive animation. Extensive experiments demonstrate that DreamWaltz-G is highly effective in generating and animating 3D avatars, outperforming existing methods in both visual quality and animation expressiveness. Our framework further supports diverse applications, including human video reenactment and multi-subject scene composition.

Bin Ren, Yawei Li, Nancy Mehta et al.

This paper provides a comprehensive review of the NTIRE 2024 challenge, focusing on efficient single-image super-resolution (ESR) solutions and their outcomes. The task of this challenge is to super-resolve an input image with a magnification factor of x4 based on pairs of low and corresponding high-resolution images. The primary objective is to develop networks that optimize various aspects such as runtime, parameters, and FLOPs, while still maintaining a peak signal-to-noise ratio (PSNR) of approximately 26.90 dB on the DIV2K_LSDIR_valid dataset and 26.99 dB on the DIV2K_LSDIR_test dataset. In addition, this challenge has 4 tracks including the main track (overall performance), sub-track 1 (runtime), sub-track 2 (FLOPs), and sub-track 3 (parameters). In the main track, all three metrics (ie runtime, FLOPs, and parameter count) were considered. The ranking of the main track is calculated based on a weighted sum-up of the scores of all other sub-tracks. In sub-track 1, the practical runtime performance of the submissions was evaluated, and the corresponding score was used to determine the ranking. In sub-track 2, the number of FLOPs was considered. The score calculated based on the corresponding FLOPs was used to determine the ranking. In sub-track 3, the number of parameters was considered. The score calculated based on the corresponding parameters was used to determine the ranking. RLFN is set as the baseline for efficiency measurement. The challenge had 262 registered participants, and 34 teams made valid submissions. They gauge the state-of-the-art in efficient single-image super-resolution. To facilitate the reproducibility of the challenge and enable other researchers to build upon these findings, the code and the pre-trained model of validated solutions are made publicly available at https://github.com/Amazingren/NTIRE2024_ESR/.

Xun Guo, Mingwu Zheng, Liang Hou et al.

Text-guided image-to-video (I2V) generation aims to generate a coherent video that preserves the identity of the input image and semantically aligns with the input prompt. Existing methods typically augment pretrained text-to-video (T2V) models by either concatenating the image with noised video frames channel-wise before being fed into the model or injecting the image embedding produced by pretrained image encoders in cross-attention modules. However, the former approach often necessitates altering the fundamental weights of pretrained T2V models, thus restricting the model's compatibility within the open-source communities and disrupting the model's prior knowledge. Meanwhile, the latter typically fails to preserve the identity of the input image. We present I2V-Adapter to overcome such limitations. I2V-Adapter adeptly propagates the unnoised input image to subsequent noised frames through a cross-frame attention mechanism, maintaining the identity of the input image without any changes to the pretrained T2V model. Notably, I2V-Adapter only introduces a few trainable parameters, significantly alleviating the training cost and also ensures compatibility with existing community-driven personalized models and control tools. Moreover, we propose a novel Frame Similarity Prior to balance the motion amplitude and the stability of generated videos through two adjustable control coefficients. Our experimental results demonstrate that I2V-Adapter is capable of producing high-quality videos. This performance, coupled with its agility and adaptability, represents a substantial advancement in the field of I2V, particularly for personalized and controllable applications.

Jian Yang, Dacheng Yin, Xiaoxuan He et al.

Recent progress in multimodal large language models (MLLMs) has highlighted the challenge of efficiently bridging pre-trained Vision-Language Models (VLMs) with Diffusion Models. While methods using a fixed number of learnable query tokens offer computational efficiency, they suffer from task generalization collapse, failing to adapt to new tasks that are distant from their pre-training tasks. To overcome this, we propose Noisy Query Tokens, which learn a distributed representation space between the VLM and Diffusion Model via end-to-end optimization, enhancing continual learning. Additionally, we introduce a VAE branch with linear projection to recover fine-grained image details. Experimental results confirm our approach mitigates generalization collapse and enables stable continual learning across diverse tasks.

Beichen Zhang, Liang Li, Zheng-Jun Zha et al.

Active learning (AL) is designed to construct a high-quality labeled dataset by iteratively selecting the most informative samples. Such sampling heavily relies on data representation, while recently pre-training is popular for robust feature learning. However, as pre-training utilizes low-level pretext tasks that lack annotation, directly using pre-trained representation in AL is inadequate for determining the sampling score. To address this problem, we propose a downstream-pretext domain knowledge traceback (DOKT) method that traces the data interactions of downstream knowledge and pre-training guidance for selecting diverse and instructive samples near the decision boundary. DOKT consists of a traceback diversity indicator and a domain-based uncertainty estimator. The diversity indicator constructs two feature spaces based on the pre-training pretext model and the downstream knowledge from annotation, by which it locates the neighbors of unlabeled data from the downstream space in the pretext space to explore the interaction of samples. With this mechanism, DOKT unifies the data relations of low-level and high-level representations to estimate traceback diversity. Next, in the uncertainty estimator, domain mixing is designed to enforce perceptual perturbing to unlabeled samples with similar visual patches in the pretext space. Then the divergence of perturbed samples is measured to estimate the domain uncertainty. As a result, DOKT selects the most diverse and important samples based on these two modules. The experiments conducted on ten datasets show that our model outperforms other state-of-the-art methods and generalizes well to various application scenarios such as semantic segmentation and image captioning.

ZhanFeng Feng, Long Peng, Xin Di et al.

Multi-frame video enhancement tasks aim to improve the spatial and temporal resolution and quality of video sequences by leveraging temporal information from multiple frames, which are widely used in streaming video processing, surveillance, and generation. Although numerous Transformer-based enhancement methods have achieved impressive performance, their computational and memory demands hinder deployment on edge devices. Quantization offers a practical solution by reducing the bit-width of weights and activations to improve efficiency. However, directly applying existing quantization methods to video enhancement tasks often leads to significant performance degradation and loss of fine details. This stems from two limitations: (a) inability to allocate varying representational capacity across frames, which results in suboptimal dynamic range adaptation; (b) over-reliance on full-precision teachers, which limits the learning of low-bit student models. To tackle these challenges, we propose a novel quantization method for video enhancement: Progressive Multi-Frame Quantization for Video Enhancement (PMQ-VE). This framework features a coarse-to-fine two-stage process: Backtracking-based Multi-Frame Quantization (BMFQ) and Progressive Multi-Teacher Distillation (PMTD). BMFQ utilizes a percentile-based initialization and iterative search with pruning and backtracking for robust clipping bounds. PMTD employs a progressive distillation strategy with both full-precision and multiple high-bit (INT) teachers to enhance low-bit models' capacity and quality. Extensive experiments demonstrate that our method outperforms existing approaches, achieving state-of-the-art performance across multiple tasks and benchmarks.The code will be made publicly available at: https://github.com/xiaoBIGfeng/PMQ-VE.

Pingyu Wu, Kai Zhu, Yu Liu et al.

Autoregressive image generation aims to predict the next token based on previous ones. However, this process is challenged by the bidirectional dependencies inherent in conventional image tokenizations, which creates a fundamental misalignment with the unidirectional nature of autoregressive models. To resolve this, we introduce AliTok, a novel Aligned Tokenizer that alters the dependency structure of the token sequence. AliTok employs a bidirectional encoder constrained by a causal decoder, a design that compels the encoder to produce a token sequence with both semantic richness and forward-dependency. Furthermore, by incorporating prefix tokens and employing a two-stage tokenizer training process to enhance reconstruction performance, AliTok achieves high fidelity and predictability simultaneously. Building upon AliTok, a standard decoder-only autoregressive model with just 177M parameters achieves a gFID of 1.44 and an IS of 319.5 on the ImageNet-256 benchmark. Scaling up to 662M parameters, our model reaches a gFID of 1.28, surpassing the state-of-the-art diffusion method while achieving a 10x faster sampling speed. The code and weights are available at https://github.com/ali-vilab/alitok.

Song Yan, Min Li, Bi Xinliang et al.

In text-driven content generation (T2C) diffusion model, semantic of generated content is mostly attributed to the process of text embedding and attention mechanism interaction. The initial noise of the generation process is typically characterized as a random element that contributes to the diversity of the generated content. Contrary to this view, this paper reveals that beneath the random surface of noise lies strong analyzable patterns. Specifically, this paper first conducts a comprehensive analysis of the impact of random noise on the model's generation. We found that noise not only contains rich semantic information, but also allows for the erasure of unwanted semantics from it in an extremely simple way based on information theory, and using the equivalence between the generation process of diffusion model and semantic injection to inject semantics into the cleaned noise. Then, we mathematically decipher these observations and propose a simple but efficient training-free and universal two-step "Semantic Erasure-Injection" process to modulate the initial noise in T2C diffusion model. Experimental results demonstrate that our method is consistently effective across various T2C models based on both DiT and UNet architectures and presents a novel perspective for optimizing the generation of diffusion model, providing a universal tool for consistent generation.

Zhipeng Huang, Shaobin Zhuang, Canmiao Fu et al.

Existing multimodal generative models fall short as qualified design copilots, as they often struggle to generate imaginative outputs once instructions are less detailed or lack the ability to maintain consistency with the provided references. In this work, we introduce WeGen, a model that unifies multimodal generation and understanding, and promotes their interplay in iterative generation. It can generate diverse results with high creativity for less detailed instructions. And it can progressively refine prior generation results or integrating specific contents from references following the instructions in its chat with users. During this process, it is capable of preserving consistency in the parts that the user is already satisfied with. To this end, we curate a large-scale dataset, extracted from Internet videos, containing rich object dynamics and auto-labeled dynamics descriptions by advanced foundation models to date. These two information are interleaved into a single sequence to enable WeGen to learn consistency-aware generation where the specified dynamics are generated while the consistency of unspecified content is preserved aligned with instructions. Besides, we introduce a prompt self-rewriting mechanism to enhance generation diversity. Extensive experiments demonstrate the effectiveness of unifying multimodal understanding and generation in WeGen and show it achieves state-of-the-art performance across various visual generation benchmarks. These also demonstrate the potential of WeGen as a user-friendly design copilot as desired. The code and models will be available at https://github.com/hzphzp/WeGen.

Hongchen Luo, Wei Zhai, Jiao Wang et al.

Perceiving potential ``action possibilities'' (\ie, affordance) regions of images and learning interactive functionalities of objects from human demonstration is a challenging task due to the diversity of human-object interactions. Prevailing affordance learning algorithms often adopt the label assignment paradigm and presume that there is a unique relationship between functional region and affordance label, yielding poor performance when adapting to unseen environments with large appearance variations. In this paper, we propose to leverage interactive affinity for affordance learning, \ie extracting interactive affinity from human-object interaction and transferring it to non-interactive objects. Interactive affinity, which represents the contacts between different parts of the human body and local regions of the target object, can provide inherent cues of interconnectivity between humans and objects, thereby reducing the ambiguity of the perceived action possibilities. To this end, we propose a visual-geometric collaborative guided affordance learning network that incorporates visual and geometric cues to excavate interactive affinity from human-object interactions jointly. Besides, a contact-driven affordance learning (CAL) dataset is constructed by collecting and labeling over 55,047 images. Experimental results demonstrate that our method outperforms the representative models regarding objective metrics and visual quality. Project: \href{https://github.com/lhc1224/VCR-Net}{github.com/lhc1224/VCR-Net}.

Jiaying Zhu, Yurui Zhu, Xin Lu et al.

Multimodal Large Language Models (MLLMs) encounter significant computational and memory bottlenecks from the massive number of visual tokens generated by high-resolution images or multi-image inputs. Previous token compression techniques are often constrained by heuristic rules that risk discarding critical information. They may suffer from biases, such as attention sinks, that lead to sharp performance drops under aggressive compression ratios. To address these limitations, we reformulate token compression as a lightweight plug-and-play framework that reformulates token compression into an end-to-end learnable decision process. To be specific, we propose VisionSelector, a scorer module decoupled from the MLLM backbone that incorporates a differentiable Top-K mechanism and a curriculum annealing strategy to bridge the training-inference gap, enabling efficient and adaptive token selection various arbitrary compression rates. Remarkably lightweight with only 12.85M trainable parameters, VisionSelector demonstrates generalization across various compression rates and adaptively identifying critical tokens. This leads to superior performance across all compression budgets, evidenced by preserving 100% accuracy on MME with 30% retention budget, outperforming prior methods by 12.14% at 10% retention budget, and doubling prefill speed. Our code is available at https://github.com/JulietChoo/VisionSelector .

Senyan Xu, Zhijing Sun, Jiaying Zhu et al.

Hybrid Event-Based Vision Sensor (HybridEVS) is a novel sensor integrating traditional frame-based and event-based sensors, offering substantial benefits for applications requiring low-light, high dynamic range, and low-latency environments, such as smartphones and wearable devices. Despite its potential, the lack of Image signal processing (ISP) pipeline specifically designed for HybridEVS poses a significant challenge. To address this challenge, in this study, we propose a coarse-to-fine framework named DemosaicFormer which comprises coarse demosaicing and pixel correction. Coarse demosaicing network is designed to produce a preliminary high-quality estimate of the RGB image from the HybridEVS raw data while the pixel correction network enhances the performance of image restoration and mitigates the impact of defective pixels. Our key innovation is the design of a Multi-Scale Gating Module (MSGM) applying the integration of cross-scale features, which allows feature information to flow between different scales. Additionally, the adoption of progressive training and data augmentation strategies further improves model's robustness and effectiveness. Experimental results show superior performance against the existing methods both qualitatively and visually, and our DemosaicFormer achieves the best performance in terms of all the evaluation metrics in the MIPI 2024 challenge on Demosaic for Hybridevs Camera. The code is available at https://github.com/QUEAHREN/DemosaicFormer.

Zhipeng Huang, Zhizheng Zhang, Yiting Lu et al.

At present, large multimodal models (LMMs) have exhibited impressive generalization capabilities in understanding and generating visual signals. However, they currently still lack sufficient capability to perceive low-level visual quality akin to human perception. Can LMMs achieve this and show the same degree of generalization in this regard? If so, not only could the versatility of LMMs be further enhanced, but also the challenge of poor cross-dataset performance in the field of visual quality assessment could be addressed. In this paper, we explore this question and provide the answer "Yes!". As the result of this initial exploration, we present VisualCritic, the first LMM for broad-spectrum image subjective quality assessment. VisualCritic can be used across diverse data right out of box, without any requirements of dataset-specific adaptation operations like conventional specialist models. As an instruction-following LMM, VisualCritic enables new capabilities of (1) quantitatively measuring the perceptual quality of given images in terms of their Mean Opinion Score (MOS), noisiness, colorfulness, sharpness, and other numerical indicators, (2) qualitatively evaluating visual quality and providing explainable descriptions, (3) discerning whether a given image is AI-generated or photographic. Extensive experiments demonstrate the efficacy of VisualCritic by comparing it with other open-source LMMs and conventional specialist models over both AI-generated and photographic images.

Kecheng Zheng, Jiawei Liu, Wei Wu et al.

Existing disentangled-based methods for generalizable person re-identification aim at directly disentangling person representations into domain-relevant interference and identity-relevant feature. However, they ignore that some crucial characteristics are stubbornly entwined in both the domain-relevant interference and identity-relevant feature, which are intractable to decompose in an unsupervised manner. In this paper, we propose a simple yet effective Calibrated Feature Decomposition (CFD) module that focuses on improving the generalization capacity for person re-identification through a more judicious feature decomposition and reinforcement strategy. Specifically, a calibrated-and-standardized Batch normalization (CSBN) is designed to learn calibrated person representation by jointly exploring intra-domain calibration and inter-domain standardization of multi-source domain features. CSBN restricts instance-level inconsistency of feature distribution for each domain and captures intrinsic domain-level specific statistics. The calibrated person representation is subtly decomposed into the identity-relevant feature, domain feature, and the remaining entangled one. For enhancing the generalization ability and ensuring high discrimination of the identity-relevant feature, a calibrated instance normalization (CIN) is introduced to enforce discriminative id-relevant information, and filter out id-irrelevant information, and meanwhile the rich complementary clues from the remaining entangled feature are further employed to strengthen it. Extensive experiments demonstrate the strong generalization capability of our framework. Our models empowered by CFD modules significantly outperform the state-of-the-art domain generalization approaches on multiple widely-used benchmarks. Code will be made public: https://github.com/zkcys001/CFD.

Yaya Shi, Xu Yang, Haiyang Xu et al.

Current metrics for video captioning are mostly based on the text-level comparison between reference and candidate captions. However, they have some insuperable drawbacks, e.g., they cannot handle videos without references, and they may result in biased evaluation due to the one-to-many nature of video-to-text and the neglect of visual relevance. From the human evaluator's viewpoint, a high-quality caption should be consistent with the provided video, but not necessarily be similar to the reference in literal or semantics. Inspired by human evaluation, we propose EMScore (Embedding Matching-based score), a novel reference-free metric for video captioning, which directly measures similarity between video and candidate captions. Benefit from the recent development of large-scale pre-training models, we exploit a well pre-trained vision-language model to extract visual and linguistic embeddings for computing EMScore. Specifically, EMScore combines matching scores of both coarse-grained (video and caption) and fine-grained (frames and words) levels, which takes the overall understanding and detailed characteristics of the video into account. Furthermore, considering the potential information gain, EMScore can be flexibly extended to the conditions where human-labeled references are available. Last but not least, we collect VATEX-EVAL and ActivityNet-FOIl datasets to systematically evaluate the existing metrics. VATEX-EVAL experiments demonstrate that EMScore has higher human correlation and lower reference dependency. ActivityNet-FOIL experiment verifies that EMScore can effectively identify "hallucinating" captions. The datasets will be released to facilitate the development of video captioning metrics. The code is available at: https://github.com/ShiYaya/emscore.

Yucheng Zhao, Guangting Wang, Chuanxin Tang et al.

Convolutional neural networks (CNN) are the dominant deep neural network (DNN) architecture for computer vision. Recently, Transformer and multi-layer perceptron (MLP)-based models, such as Vision Transformer and MLP-Mixer, started to lead new trends as they showed promising results in the ImageNet classification task. In this paper, we conduct empirical studies on these DNN structures and try to understand their respective pros and cons. To ensure a fair comparison, we first develop a unified framework called SPACH which adopts separate modules for spatial and channel processing. Our experiments under the SPACH framework reveal that all structures can achieve competitive performance at a moderate scale. However, they demonstrate distinctive behaviors when the network size scales up. Based on our findings, we propose two hybrid models using convolution and Transformer modules. The resulting Hybrid-MS-S+ model achieves 83.9% top-1 accuracy with 63M parameters and 12.3G FLOPS. It is already on par with the SOTA models with sophisticated designs. The code and models are publicly available at https://github.com/microsoft/SPACH.

Zehui Chen, Chenhongyi Yang, Qiaofei Li et al.

Deep learning-based dense object detectors have achieved great success in the past few years and have been applied to numerous multimedia applications such as video understanding. However, the current training pipeline for dense detectors is compromised to lots of conjunctions that may not hold. In this paper, we investigate three such important conjunctions: 1) only samples assigned as positive in classification head are used to train the regression head; 2) classification and regression share the same input feature and computational fields defined by the parallel head architecture; and 3) samples distributed in different feature pyramid layers are treated equally when computing the loss. We first carry out a series of pilot experiments to show disentangling such conjunctions can lead to persistent performance improvement. Then, based on these findings, we propose Disentangled Dense Object Detector (DDOD), in which simple and effective disentanglement mechanisms are designed and integrated into the current state-of-the-art dense object detectors. Extensive experiments on MS COCO benchmark show that our approach can lead to 2.0 mAP, 2.4 mAP and 2.2 mAP absolute improvements on RetinaNet, FCOS, and ATSS baselines with negligible extra overhead. Notably, our best model reaches 55.0 mAP on the COCO test-dev set and 93.5 AP on the hard subset of WIDER FACE, achieving new state-of-the-art performance on these two competitive benchmarks. Code is available at https://github.com/zehuichen123/DDOD.

Jiawei Liu, Zhipeng Huang, Kecheng Zheng et al.

Although existing person re-identification (Re-ID) methods have shown impressive accuracy, most of them usually suffer from poor generalization on unseen target domain. Thus, generalizable person Re-ID has recently drawn increasing attention, which trains a model on source domains that generalizes well on unseen target domain without model updating. In this work, we propose a novel adaptive domain-specific normalization approach (AdsNorm) for generalizable person Re-ID. It describes unseen target domain as a combination of the known source ones, and explicitly learns domain-specific representation with target distribution to improve the model's generalization by a meta-learning pipeline. Specifically, AdsNorm utilizes batch normalization layers to collect individual source domains' characteristics, and maps source domains into a shared latent space by using these characteristics, where the domain relevance is measured by a distance function of different domain-specific normalization statistics and features. At the testing stage, AdsNorm projects images from unseen target domain into the same latent space, and adaptively integrates the domain-specific features carrying the source distributions by domain relevance for learning more generalizable aggregated representation on unseen target domain. Considering that target domain is unavailable during training, a meta-learning algorithm combined with a customized relation loss is proposed to optimize an effective and efficient ensemble model. Extensive experiments demonstrate that AdsNorm outperforms the state-of-the-art methods. The code is available at: https://github.com/hzphzp/AdsNorm.

Shaofei Cai, Liang Li, Jincan Deng et al.

Graph neural networks (GNNs) emerged recently as a standard toolkit for learning from data on graphs. Current GNN designing works depend on immense human expertise to explore different message-passing mechanisms, and require manual enumeration to determine the proper message-passing depth. Inspired by the strong searching capability of neural architecture search (NAS) in CNN, this paper proposes Graph Neural Architecture Search (GNAS) with novel-designed search space. The GNAS can automatically learn better architecture with the optimal depth of message passing on the graph. Specifically, we design Graph Neural Architecture Paradigm (GAP) with tree-topology computation procedure and two types of fine-grained atomic operations (feature filtering and neighbor aggregation) from message-passing mechanism to construct powerful graph network search space. Feature filtering performs adaptive feature selection, and neighbor aggregation captures structural information and calculates neighbors' statistics. Experiments show that our GNAS can search for better GNNs with multiple message-passing mechanisms and optimal message-passing depth. The searched network achieves remarkable improvement over state-of-the-art manual designed and search-based GNNs on five large-scale datasets at three classical graph tasks. Codes can be found at https://github.com/phython96/GNAS-MP.

Jing Zhang, Yang Cao, Zheng-Jun Zha et al.

Increasing the visibility of nighttime hazy images is challenging because of uneven illumination from active artificial light sources and haze absorbing/scattering. The absence of large-scale benchmark datasets hampers progress in this area. To address this issue, we propose a novel synthetic method called 3R to simulate nighttime hazy images from daytime clear images, which first reconstructs the scene geometry, then simulates the light rays and object reflectance, and finally renders the haze effects. Based on it, we generate realistic nighttime hazy images by sampling real-world light colors from a prior empirical distribution. Experiments on the synthetic benchmark show that the degrading factors jointly reduce the image quality. To address this issue, we propose an optimal-scale maximum reflectance prior to disentangle the color correction from haze removal and address them sequentially. Besides, we also devise a simple but effective learning-based baseline which has an encoder-decoder structure based on the MobileNet-v2 backbone. Experiment results demonstrate their superiority over state-of-the-art methods in terms of both image quality and runtime. Both the dataset and source code will be available at https://github.com/chaimi2013/3R.

Ganchao Tan, Daqing Liu, Meng Wang et al.

Generating natural language descriptions for videos, i.e., video captioning, essentially requires step-by-step reasoning along the generation process. For example, to generate the sentence "a man is shooting a basketball", we need to first locate and describe the subject "man", next reason out the man is "shooting", then describe the object "basketball" of shooting. However, existing visual reasoning methods designed for visual question answering are not appropriate to video captioning, for it requires more complex visual reasoning on videos over both space and time, and dynamic module composition along the generation process. In this paper, we propose a novel visual reasoning approach for video captioning, named Reasoning Module Networks (RMN), to equip the existing encoder-decoder framework with the above reasoning capacity. Specifically, our RMN employs 1) three sophisticated spatio-temporal reasoning modules, and 2) a dynamic and discrete module selector trained by a linguistic loss with a Gumbel approximation. Extensive experiments on MSVD and MSR-VTT datasets demonstrate the proposed RMN outperforms the state-of-the-art methods while providing an explicit and explainable generation process. Our code is available at https://github.com/tgc1997/RMN.

Yuxin Wang, Hongtao Xie, Zhengjun Zha et al.

Scene text detection has witnessed rapid development in recent years. However, there still exists two main challenges: 1) many methods suffer from false positives in their text representations; 2) the large scale variance of scene texts makes it hard for network to learn samples. In this paper, we propose the ContourNet, which effectively handles these two problems taking a further step toward accurate arbitrary-shaped text detection. At first, a scale-insensitive Adaptive Region Proposal Network (Adaptive-RPN) is proposed to generate text proposals by only focusing on the Intersection over Union (IoU) values between predicted and ground-truth bounding boxes. Then a novel Local Orthogonal Texture-aware Module (LOTM) models the local texture information of proposal features in two orthogonal directions and represents text region with a set of contour points. Considering that the strong unidirectional or weakly orthogonal activation is usually caused by the monotonous texture characteristic of false-positive patterns (e.g. streaks.), our method effectively suppresses these false positives by only outputting predictions with high response value in both orthogonal directions. This gives more accurate description of text regions. Extensive experiments on three challenging datasets (Total-Text, CTW1500 and ICDAR2015) verify that our method achieves the state-of-the-art performance. Code is available at https://github.com/wangyuxin87/ContourNet.