Video Large Language Models (video LLMs) have demonstrated remarkable capabilities in video understanding tasks, such as video question answering and temporal localization. However, understanding long videos still remains a significant challenge. Existing video LLMs adopt uni-granularity tokens for long videos, failing to simultaneously understand both high-level semantics and low-level visual details in videos. To tackle this problem, we propose ReCrossVLLM, a reflective cross-granularity grounding framework for video LLM with preference optimization to collaboratively achieve long video understanding, which not only retains the capabilities of high-level video semantics understanding, but also strengthens the fine-grained understanding abilities. Specifically, we propose the coarse-to-fine grounding and fine-to-coarse reflection strategies for long video understanding. In the coarse-to-fine grounding strategy, the video LLM with a coarse-grained module first locates the key video segments from the long video by tackling massive frames of the long video with fewer per-frame tokens. And then video LLM adapted with the fine-grained module further analyzes the key video segments with more per-frame tokens so that it can understand fine-grained information. In case the video LLM locates the wrong key video segments, during the inference stage, our designed fine-to-coarse reflection strategy instructs the fine-grained module to reflect the effectiveness of the locating result and decide whether to return to the coarse-to-fine grounding strategy with reflection feedback. Additionally, during the training stage, the coarse-to-fine grounding strategy is optimized with our proposed cross-granularity preference optimization strategy to further improve grounding efficiency. Extensive experiments for long video question answering and temporal video grounding tasks demonstrate that our proposed ReCrossVLLM framework can significantly improve the Video Large Language Model for long video understanding.

Neural radiance fields, which represent a 3D scene as a color field and a density field, have demonstrated great progress in novel view synthesis yet are unfavorable for editing due to the implicitness. This work studies the task of efficient 3D editing, where we focus on editing speed and user interactivity. To this end, we propose to learn the color field as an explicit 2D appearance aggregation, also called canonical image, with which users can easily customize their 3D editing via 2D image processing. We complement the canonical image with a projection field that maps 3D points onto 2D pixels for texture query. This field is initialized with a pseudo canonical camera model and optimized with offset regularity to ensure the naturalness of the canonical image. Extensive experiments on different datasets suggest that our representation, dubbed AGAP, well supports various ways of 3D editing (e.g., stylization, instance segmentation, and interactive drawing). Our approach demonstrates remarkable efficiency by being at least 20× faster per edit compared to existing NeRF-based editing methods. Project page is available at h ttps: //felixcheng97.github.io/AGAP/.

Due to the difficulty in scaling up, generative adversarial networks (GANs) seem to be falling out of grace with the task of text-conditioned image synthesis. Sparsely activated mixture-of-experts (MoE) has recently been demonstrated as a valid solution to training large-scale models with limited resources. Inspired by this, we present Aurora, a GAN-based text-to-image generator that employs a collection of experts to learn feature processing, together with a sparse router to adaptively select the most suitable expert for each feature point. We adopt a two-stage training strategy, which first learns a base model at 64 × 64 resolution followed by an upsampler to produce 512 × 512 images. Trained with only public data, our approach encouragingly closes the performance gap between GANs and industry-level diffusion models, maintaining a fast inference speed. We release the code and checkpoints here to facilitate the community for further development.

Generating detailed captions comprehending text-rich visual content in images has received growing attention for Large Vision-Language Models (LVLMs). However, few studies have developed benchmarks specifically tailored for detailed captions to measure their accuracy and comprehensiveness. In this paper, we introduce a detailed caption benchmark, termed as CompreCap, to evaluate the visual context from a directed scene graph view. Concretely, we first manually segment the image into semantically meaningful regions (i.e., semantic segmentation mask) according to common-object vocabulary, while also distinguishing attributes of objects within all those regions. Then directional relation labels of these objects are annotated to compose a directed scene graph that can well encode rich compositional information of the image. Based on our directed scene graph, we develop a pipeline to assess the generated detailed captions from LVLMs on multiple levels, including the object-level coverage, the accuracy of attribute descriptions, the score of key relationships, etc. Experimental results on the CompreCap dataset confirm that our evaluation method aligns closely with human evaluation scores across LVLMs. We have released the code and the dataset here to support the community.

Although text-to-image (T2I) models have recently thrived as visual generative priors, their reliance on high-quality text-image pairs makes scaling up expensive. We argue that grasping the cross-modality alignment is not a necessity for a sound visual generative prior, whose focus should be on texture modeling. Such a philosophy inspires us to study image-to-image (I2I) generation, where models can learn from in-the-wild images in a self-supervised manner. We first develop a pure vision-based training framework, Lumos, and confirm the feasibility and the scalability of learning I2I models. We then find that, as an upstream task of T2I, our I2I model serves as a more foundational visual prior and achieves on-par or better performance than existing T2I models using only 1/10 text-image pairs for fine-tuning. We further demonstrate the superiority of I2I priors over T2I priors on some text-irrelevant visual generative tasks, like image-to-3D and image-to-video.

The image-to-video (I2V) generation is conditioned on the static image, which has been enhanced recently by the motion intensity as an additional control signal. These motion-aware models are appealing to generate diverse motion patterns, yet there lacks a reliable motion estimator for training such models on large-scale video set in the wild. Traditional metrics, e.g., SSIM or optical flow, are hard to generalize to arbitrary videos, while, it is very tough for human annotators to label the abstract motion intensity neither. Furthermore, the motion intensity shall reveal both local object motion and global camera movement, which has not been studied before. This paper addresses the challenge with a new motion estimator, capable of measuring the decoupled motion intensities of objects and cameras in video. We leverage the contrastive learning on randomly paired videos and distinguish the video with greater motion intensity. Such a paradigm is friendly for annotation and easy to scale up to achieve stable performance on motion estimation. We then present a new I2V model, named MotionStone, developed with the decoupled motion estimator. Experimental results demonstrate the stability of the proposed motion estimator and the state-of-the-art performance of MotionStone on I2V generation. These advantages warrant the decoupled motion estimator to serve as a general plug-in enhancer for both data processing and video generation training.

Text serves as the key control signal in video generation due to its narrative nature. To render text descriptions into video clips, current video diffusion models borrow features from text encoders yet struggle with limited text comprehension. The recent success of large language models (LLMs) showcases the power of decoder-only transformers, which offers three clear benefits for text-to-video (T2V) generation, namely, precise text understanding resulting from the superior scalability, imagination beyond the input text enabled by next token prediction, and flexibility to prioritize user interests through instruction tuning. Nevertheless, the feature distribution gap emerging from the two different text modeling paradigms hinders the direct use of LLMs in established T2V models. This work addresses this challenge with Mimir, an end-to-end training framework featuring a carefully tailored token fuser to harmonize the outputs from text encoders and LLMs. Such a design allows the T2V model to fully leverage learned video priors while capitalizing on the text-related capability of LLMs. Extensive quantitative and qualitative results demonstrate the effectiveness of Mimir in generating high-quality videos with excellent text comprehension, especially when processing short captions and managing shifting motions. Project page: https://lucaria-academy.github.io/Mimir/

The Audio Description (AD) task aims to generate descriptions of visual elements for visually impaired individuals to help them access long-form video content, like movies. With video feature, text, character bank and context information as inputs, the generated ADs are able to correspond to the characters by name and provide reasonable, contextual descriptions to help audience understand the storyline of movie. To achieve this goal, we propose to leverage pre-trained foundation models through a simple and unified framework to generate ADs with interleaved multimodal sequence as input, termed as Uni-AD. To enhance the alignment of features across various modalities with finer granularity, we introduce a simple and lightweight module that maps video features into the textual feature space. Moreover, we also propose a character-refinement module to provide more precise information by identifying the main characters who play more significant roles in the video context. With these unique designs, we further incorporate contextual information and a contrastive loss into our architecture to generate smoother and more contextually appropriate ADs. Experiments on multiple AD datasets show that Uni-AD performs well on AD generation, which demonstrates the effectiveness of our approach. Our code is available at: https://github.com/ant-research/UniAD.

We propose Framer for interactive frame interpolation, which targets producing smoothly transitioning frames between two images as per user creativity. Concretely, besides taking the start and end frames as inputs, our approach supports customizing the transition process by tailoring the trajectory of some selected keypoints. Such a design enjoys two clear benefits. First, incorporating human interaction mitigates the issue arising from numerous possibilities of transforming one image to another, and in turn enables finer control of local motions. Second, as the most basic form of interaction, keypoints help establish the correspondence across frames, enhancing the model to handle challenging cases (e.g., objects on the start and end frames are of different shapes and styles). It is noteworthy that our system also offers an "autopilot" mode, where we introduce a module to estimate the keypoints and refine the trajectory automatically, to simplify the usage in practice. Extensive experimental results demonstrate the appealing performance of Framer on various applications, such as image morphing, time-lapse video generation, cartoon interpolation, etc. The code, model, and interface are publicly accessible at https://github.com/aim-uofa/Framer.

Audio is essential for multimodal video understanding. On the one hand, video inherently contains audio, which supplies complementary information to vision. Besides, video large language models (Video-LLMs) can encounter many audio-centric settings. However, existing Video-LLMs and Audio-Visual Large Language Models (AV-LLMs) exhibit deficiencies in exploiting audio information, leading to weak understanding and hallucinations. To solve the issues, we delve into the model architecture and dataset. (1) From the architectural perspective, we propose a fine-grained AV-LLM, namely Dolphin. The concurrent alignment of audio and visual modalities in both temporal and spatial dimensions ensures a comprehensive and accurate understanding of videos. Specifically, we devise an audio-visual multi-scale adapter for multi-scale information aggregation, which achieves spatial alignment. For temporal alignment, we propose audio-visual interleaved merging. (2) From the dataset perspective, we curate an audio-visual caption \& instruction-tuning dataset, called AVU. It comprises 5.2 million diverse, open-ended data tuples (video, audio, question, answer) and introduces a novel data partitioning strategy. Extensive experiments show our model not only achieves remarkable performance in audio-visual understanding, but also mitigates potential hallucinations.

Character image animation, which generates high-quality videos from a reference image and target pose sequence, has seen significant progress in recent years. However, most existing methods only apply to human figures, which usually do not generalize well on anthropomorphic characters commonly used in industries like gaming and entertainment. Our in-depth analysis suggests to attribute this limitation to their insufficient modeling of motion, which is unable to comprehend the movement pattern of the driving video, thus imposing a pose sequence rigidly onto the target character. To this end, this paper proposes $\texttt{Animate-X}$, a universal animation framework based on LDM for various character types (collectively named $\texttt{X}$), including anthropomorphic characters. To enhance motion representation, we introduce the Pose Indicator, which captures comprehensive motion pattern from the driving video through both implicit and explicit manner. The former leverages CLIP visual features of a driving video to extract its gist of motion, like the overall movement pattern and temporal relations among motions, while the latter strengthens the generalization of LDM by simulating possible inputs in advance that may arise during inference. Moreover, we introduce a new Animated Anthropomorphic Benchmark ($\texttt{$A^2$Bench}$) to evaluate the performance of $\texttt{Animate-X}$ on universal and widely applicable animation images. Extensive experiments demonstrate the superiority and effectiveness of $\texttt{Animate-X}$ compared to state-of-the-art methods.

Masked-based autoregressive models have demonstrated promising image generation capability in continuous space. However, their potential for video generation remains under-explored. Masked-based autoregressive models have demonstrated promising image generation capability in continuous space. However, their potential for video generation remains under-explored. In this paper, we propose \textbf{VideoMAR}, a concise and efficient decoder-only autoregressive image-to-video model with continuous tokens, composing temporal frame-by-frame and spatial masked generation. We first identify temporal causality and spatial bi-directionality as the first principle of video AR models, and propose the next-frame diffusion loss for the integration of mask and video generation. Besides, the huge cost and difficulty of long sequence autoregressive modeling is a basic but crucial issue. To this end, we propose the temporal short-to-long curriculum learning and spatial progressive resolution training, and employ progressive temperature strategy at inference time to mitigate the accumulation error. Furthermore, VideoMAR replicates several unique capacities of language models to video generation. It inherently bears high efficiency due to simultaneous temporal-wise KV cache and spatial-wise parallel generation, and presents the capacity of spatial and temporal extrapolation via 3D rotary embeddings. On the VBench-I2V benchmark, VideoMAR surpasses the previous state-of-the-art (Cosmos I2V) while requiring significantly fewer parameters ($9.3\%$), training data ($0.5\%$), and GPU resources ($0.2\%$).

Learning joint and coordinated features across modalities is essential for many audio-visual tasks. Existing pre-training methods primarily focus on global information, neglecting fine-grained features and positions, leading to suboptimal performance in dense prediction tasks. To address this issue, we take a further step towards region-aware audio-visual pre-training and propose CrossMAE, which excels in Cross-modality interaction and region alignment. Specifically, we devise two masked autoencoding (MAE) pretext tasks at both pixel and embedding levels, namely Cross-Conditioned Reconstruction and Cross-Embedding Reconstruction. Taking the visual modality as an example (the same goes for audio), in Cross-Conditioned Reconstruction, the visual modality reconstructs the input image pixels conditioned on audio Attentive Tokens. As for the more challenging Cross-Embedding Reconstruction, unmasked visual tokens reconstruct complete audio features under the guidance of Learnable Queries implying positional information, which effectively enhances the interaction between modalities and exploits fine-grained semantics. Experimental results demonstrate that CrossMAE achieves state-of-the-art performance not only in classification and retrieval, but also in dense prediction tasks. Furthermore, we dive into the mechanism of modal interaction and region alignment of CrossMAE, highlighting the effectiveness of the proposed components.

We present the content deformation field (CoDeF) as a new type of video representation, which consists of a canonical content field aggregating the static contents in the entire video and a temporal deformation field recording the transformations from the canonical image (i.e., rendered from the canonical content field) to each individual frame along the time axis. Given a target video, these two fields are jointly optimized to reconstruct it through a carefully tailored rendering pipeline. We advisedly introduce some regularizations into the optimization process, urging the canonical content field to inherit semantics (e.g., the object shape) from the video. With such a design, CoDeF naturally supports lifting image algorithms for video processing, in the sense that one can apply an image algorithm to the canonical image and effortlessly propagate the outcomes to the entire video with the aid of the temporal deformation field. We experimentally show that CoDeF is able to lift image-to-image translation to video-to-video translation and lift keypoint detection to keypoint tracking without any training. More importantly, thanks to our lifting strategy that deploys the algorithms on only one image, we achieve superior cross-frame consistency in processed videos compared to existing video-to-video translation approaches, and even manage to track non-rigid objects like water and smog. Code is made available at https: / /qiuyu96. github.io/CoDeF/

This work presents a unified knowledge protocol, called UKnow, which facilitates knowledge-based studies from the perspective of data. Particularly focusing on visual and linguistic modalities, we categorize data knowledge into five unit types, namely, in-image, in-text, cross-image, cross-text, and image-text, and set up an efficient pipeline to help construct the multimodal knowledge graph from any data collection. Thanks to the logical information naturally contained in knowledge graph, organizing datasets under UKnow format opens up more possibilities of data usage compared to the commonly used image-text pairs. Following UKnow protocol, we collect, from public international news, a large-scale multimodal knowledge graph dataset that consists of 1,388,568 nodes (with 571,791 vision-related ones) and 3,673,817 triplets. The dataset is also annotated with rich event tags, including 11 coarse labels and 9,185 fine labels. Experiments on four benchmarks demonstrate the potential of UKnow in supporting common-sense reasoning and boosting vision-language pre-training with a single dataset, benefiting from its unified form of knowledge organization. Code, dataset, and models will be made publicly available. See Appendix to download the dataset.

This paper introduces Chain-of-Sight, a vision-language bridge module that accelerates the pre-training of Multimodal Large Language Models (MLLMs). Our approach employs a sequence of visual resamplers that capture visual details at various spacial scales. This architecture not only leverages global and local visual contexts effectively, but also facilitates the flexible extension of visual tokens through a compound token scaling strategy, allowing up to a 16x increase in the token count post pre-training. Consequently, Chain-of-Sight requires significantly fewer visual tokens in the pre-training phase compared to the fine-tuning phase. This intentional reduction of visual tokens during pre-training notably accelerates the pre-training process, cutting down the wall-clock training time by $\sim$73\%. Empirical results on a series of vision-language benchmarks reveal that the pre-train acceleration through Chain-of-Sight is achieved without sacrificing performance, matching or surpassing the standard pipeline of utilizing all visual tokens throughout the entire training process. Further scaling up the number of visual tokens for pre-training leads to stronger performances, competitive to existing approaches in a series of benchmarks.

In this work, we empirically confirm that the key reason causing such an issue is that the training images are usually paired with short captions, leaving certain tokens easily overshadowed by salient tokens. Towards this problem, our initial attempt is to relabel the data with long captions, however, directly learning with which may lead to performance degradation in understanding short text (e.g., in the image classification task). Then, after incorporating corner tokens to aggregate diverse textual information, we manage to help the model catch up to its original level of short text understanding yet greatly enhance its capability of long text understanding. We further look into whether the model can continuously benefit from longer captions and notice a clear trade-off between the performance and the efficiency. Finally, we validate the effectiveness of our approach using a self-constructed large-scale dataset, which consists of 100M long caption oriented text-image pairs. Our method achieves superior performance in long-text-image retrieval tasks. The project page is available at https://wuw2019.github.io/lot-lip.

Source-Free Domain Adaptation (SFDA) task aims to transfer knowledge from a labeled source domain to a label-scarce target domain, in which the source data can not be accessed but only a pre-trained source model and unlabeled target data are available during adaptation. Previous methods for source model adaptation rely on hypothesis transfer learning that trains the feature extractor to learn target features aligned to the distribution of source features while freezing the source classifier. However, reusing only the source classifier without exploring the comprehensive knowledge of the source model can lead to biased feature alignment. To this end, we propose a novel method called Transformer-bAsed thorouGh Source HypOthesis Transfer (TagSHOT) framework to effectively unleash the thorough knowledge potential of pre-trained source hypothesis. Specifically, our approach delves into the correlation coefficient among CLS/patch tokens across different Transformer layers, uncovering the concealed insights within the pre-trained source model and constructing a comprehensive source hypothesis. By tailoring the target feature alignment to this thorough source hypothesis, our model facilitates the adaptation of a broader range of classification-related knowledge to the target domain. Furthermore, we introduce a Salient Token Extension (STE) module, designed to capture the target-specific discriminative information by propagating the salient information among tokens. This mechanism enriches our model's ability to understand and incorporate target-specific nuances. Extensive experiments have been conducted to validate the effectiveness of our method, which outperforms state-of-the-art approaches by a large margin.

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 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 to reproduce the results in this paper is available at https://github.com/RuiLiFengiNeural-Dependencies.

Semantically coherent out-of-distribution (SCOOD) detection aims to discern outliers from the intended data distribution with access to unlabeled extra set. The coexistence of in-distribution and out-of-distribution samples will exacerbate the model overfitting when no distinction is made. To address this problem, we propose a novel uncertainty-aware optimal transport scheme. Our scheme consists of an energy-based transport (ET) mechanism that estimates the fluctuating cost of uncertainty to promote the assignment of semantic-agnostic representation, and an inter-cluster extension strategy that enhances the discrimination of semantic property among different clusters by widening the corresponding margin distance. Furthermore, a T-energy score is presented to mitigate the magnitude gap between the parallel transport and classifier branches. Extensive experiments on two standard SCOOD benchmarks demonstrate the above-par OOD detection performance, outperforming the state-of-the-art methods by a margin of 27.69% and 34.4% on FPR@95, respectively. Code is available at https://github.com/LuFan3/IET-OOD.

Non-exemplar class-incremental learning aims to recognize both the old and new classes without access to old class samples. The conflict between old and new class optimization is exacerbated since the shared neural pathways can only be differentiated by the incremental samples. To address this problem, we propose a novel self-organizing pathway expansion scheme. Our scheme consists of a class-specific pathway organization strategy that reduces the coupling of optimization pathway among different classes to enhance the independence of the feature representation, and a pathway-guided feature optimization mechanism to mitigate the update interference between the old and new classes. Extensive experiments on four datasets demonstrate significant performance gains, outperforming the state-of-the-art methods by a margin of 1%, 3%, 2% and 2%, respectively.

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.

Human brains respond to semantic features of presented stimuli with different neurons. This raises the question of whether deep neural networks admit a similar behavior pattern. To investigate this phenomenon, this paper identifies a small cluster of neurons associated with a specific subject in a diffusion model. We call those neurons the concept neurons. They can be identified by statistics of network gradients to a stimulation connected with the given subject. The concept neurons demonstrate magnetic properties in interpreting and manipulating generation results. Shutting them can directly yield the related subject contextualized in different scenes. Concatenating multiple clusters of concept neurons can vividly generate all related concepts in a single image. Our method attains impressive performance for multi-subject customization, even four or more subjects. For large-scale applications, the concept neurons are environmentally friendly as we only need to store a sparse cluster of int index instead of dense float32 parameter values, reducing storage consumption by 90% compared with previous customized generation methods. Extensive qualitative and quantitative studies on diverse scenarios show the superiority of our method in interpreting and manipulating diffusion models.

Vision-language representation learning models (e.g., CLIP) have achieved state-of-the-art performance on various downstream tasks, which usually need large-scale training data to learn discriminative representation. Recent progress on generative diffusion models (e.g., DALL-E 2) has demonstrated that diverse high-quality samples can be synthesized by randomly sampling from generative distribution. By virtue of generative capability in this paper, we propose a novel vision-language Representation Learning method with diffusion-based Embedding Generation (RLEG), which exploits diffusion models to generate feature embedding online for learning effective vision-language representation. Specifically, we first adopt image and text encoders to extract the corresponding embeddings. Secondly, pretrained diffusion-based embedding generators are harnessed to transfer the embedding modality online between vision and language domains. The embeddings generated from the generators are then served as augmented embedding-level samples, which are applied to contrastive learning with the variant of the CLIP framework. Experimental results show that the proposed method could learn effective representation and achieve state-of-the-art performance on various tasks including image classification, image-text retrieval, object detection, semantic segmentation, and text-conditional image generation.

Synthesizing images with user-specified subjects has received growing attention due to its practical applications. Despite the recent success in single subject customization, existing algorithms suffer from high training cost and low success rate along with increased number of subjects. Towards controllable image synthesis with multiple subjects as the constraints, this work studies how to efficiently represent a particular subject as well as how to appropriately compose different subjects. We find that the text embedding regarding the subject token already serves as a simple yet effective representation that supports arbitrary combinations without any model tuning. Through learning a residual on top of the base embedding, we manage to robustly shift the raw subject to the customized subject given various text conditions. We then propose to employ layout, a very abstract and easy-to-obtain prior, as the spatial guidance for subject arrangement. By rectifying the activations in the cross-attention map, the layout appoints and separates the location of different subjects in the image, significantly alleviating the interference across them. Using cross-attention map as the intermediary, we could strengthen the signal of target subjects and weaken the signal of irrelevant subjects within a certain region, significantly alleviating the interference across subjects. Both qualitative and quantitative experimental results demonstrate our superiority over state-of-the-art alternatives under a variety of settings for multi-subject customization.

Despite the rapid advance of 3D-aware image synthesis, existing studies usually adopt a mixture of techniques and tricks, leaving it unclear how each part contributes to the final performance in terms of generality. Following the most popular and effective paradigm in this field, which incorporates a neural radiance field (NeRF) into the generator of a generative adversarial network (GAN), we builda well-structured codebase through modularizing the generation process. Such a design allows researchers to develop and replace each module independently, and hence offers an opportunity to fairly compare various approaches and recognize their contributions from the module perspective. The reproduction of a range of cutting-edge algorithms demonstrates the availability of our modularized codebase. We also perform a variety of in-depth analyses, such as the comparison across different types of point feature, the necessity of the tailing upsampler in the generator, the reliance on the camera pose prior, etc., which deepen our understanding of existing methods and point out some further directions of the research work. Code and models will be made publicly available to facilitate the development and evaluation of this field.

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.

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.

Cloth-Changing person re-identification (CC-ReID) aims at matching the same person across different locations over a long-duration, e.g., over days, and therefore inevitably has cases of changing clothing. In this paper, we focus on handling well the CC-ReID problem under a more challenging setting, i.e., just from a single image, which enables an efficient and latency-free person identity matching for surveillance. Specifically, we introduce Gait recognition as an auxiliary task to drive the Image ReID model to learn cloth-agnostic representations by leveraging personal unique and cloth-independent gait information, we name this framework as GI-ReID. GI-ReID adopts a two-stream architecture that consists of an image ReID-Stream and an auxiliary gait recognition stream (Gait-Stream). The Gait-Stream, that is discarded in the inference for high efficiency, acts as a regulator to encourage the ReID-Stream to capture cloth-invariant biometric motion features during the training. To get temporal continuous motion cues from a single image, we design a Gait Sequence Prediction (GSP) module for Gait-Stream to enrich gait information. Finally, a semantics consistency constraint over two streams is enforced for effective knowledge regularization. Extensive experiments on multiple image-based Cloth-Changing ReID benchmarks, e.g., LTCC, PRCC, Real28, and VC-Clothes, demonstrate that GI-ReID performs favorably against the state-of-the-art methods.

Image-to-video person re-identification aims to retrieve the same pedestrian as the image-based query from a video-based gallery set. Existing methods treat it as a cross-modality retrieval task and learn the common latent embeddings from image and video modalities, which are both less effective and efficient due to large modality gap and redundant feature learning by utilizing all video frames. In this work, we first regard this task as point-to-set matching problem identical to human decision process, and propose a novel Temporal Complementarity-Guided Reinforcement Learning (TCRL) approach for image-to-video person re-identification. TCRL employs deep reinforcement learning to make sequential judgments on dynamically selecting suitable amount of frames from gallery videos, and accumulate adequate temporal complementary information among these frames by the guidance of the query image, towards balancing efficiency and accuracy. Specifically, TCRL formulates point-to-set matching procedure as Markov decision process, where a sequential judgement agent measures the uncertainty between the query image and all historical frames at each time step, and verifies that sufficient complementary clues are accumulated for judgment (same or different) or one more frames are requested to assist judgment. Moreover, TCRL maintains a sequential feature extraction module with complementary residual detectors to dynamically suppress redundant salient regions and thoroughly mine diverse complementary clues among these selected frames for enhancing frame-level representation. Extensive experiments demonstrate the superiority of our method.

Existing person reidentification (ReID) methods typically load the pretrained ImageNet weights for initialization directly. However, as a fine-grained classification task, ReID is more challenging and there exists a large domain gap between ImageNet classification. Inspired by the great success of self-supervised representation learning with contrastive objectives, in this paper, we design an Unsupervised Pretraining framework for reidentification (UP-ReID) based on the contrastive learning (CL) pipeline. During the pre-training, we attempt to address two critical issues for learning fine-grained ReID features: (1) the augmentations in the CL pipeline usually distort the discriminative clues in person images, and (2) the fine-grained local features of person images are not fully-explored. Therefore, we introduce an intra-identity (12-) regularization in the UP-ReID, which is instantiated as two constraints coming from the global image and local patch aspects, respectively. A global consistency constraint is enforced between augmented and original person images to increase robustness to augmentation, while an intrinsic contrastive constraint among local patches of each image is employed to fully explore the local discriminative clues. Extensive experiments on multiple popular reid datasets, PersonX, Market1501, CUHK03, and MSMT17, demonstrate that our UP-ReID pretrained model can significantly benefit the downstream ReID fine-tuning and achieve state-of-the-art performance.

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.

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 led 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 monotone decreasing property of network ranks 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 ranks in realistic settings, \ieno, 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. Code to detect the rank behavior of networks can be found in https://github.com/RuiLiFeng/Rank-Diminishing-in-Deep-Neural-Networks.

Incremental implicitly-refined classification task aims at assigning hierarchical labels to each sample encountered at different phases. Existing methods tend to fail in generating hierarchy-invariant descriptors when the novel classes are inherited from the old ones. To address the issue, this paper, which explores the inheritance relations in the process of multi-level semantic increment, proposes an Uncertainty-Aware Hierarchical Refinement (UAHR) scheme. Specifically, our proposed scheme consists of a global representation extension strategy that enhances the discrimination of incremental representation by widening the corresponding margin distance, and a hierarchical distribution alignment strategy that refines the distillation process by explicitly determining the inheritance relationship of the incremental class. Particularly, the shifting subclasses are corrected under the guidance of hierarchical uncertainty, ensuring the consistency of the homogeneous features. Extensive experiments on widely used benchmarks (i.e., IIRC-CIFAR, IIRC-ImageNet-lite, IIRC-ImageNet-Subset, and IIRC-ImageNet-full) demonstrate the superiority of our proposed method over the state-of-the-art approaches.

Many unsupervised domain adaptive (UDA) person ReID approaches combine clustering-based pseudo-label prediction with feature fine-tuning. However, because of domain gap, the pseudo-labels are not always reliable and there are noisy/incorrect labels. This would mislead the feature representation learning and deteriorate the performance. In this paper, we propose to estimate and exploit the credibility of the assigned pseudo-label of each sample to alleviate the influence of noisy labels, by suppressing the contribution of noisy samples. We build our baseline framework using the mean teacher method together with an additional contrastive loss. We have observed that a sample with a wrong pseudo-label through clustering in general has a weaker consistency between the output of the mean teacher model and the student model. Based on this finding, we propose to exploit the uncertainty (measured by consistency levels) to evaluate the reliability of the pseudo-label of a sample and incorporate the uncertainty to re-weight its contribution within various ReID losses, including the ID classification loss per sample, the triplet loss, and the contrastive loss. Our uncertainty-guided optimization brings significant improvement and achieves the state-of-the-art performance on benchmark datasets.

Video-based person re-identification aims to match pedestrians from video sequences across non-overlapping camera views. The key factor for video person re-identification is to effectively exploit both spatial and temporal clues from video sequences. In this work, we propose a novel Spatial-Temporal Correlation and Topology Learning framework (CTL) to pursue discriminative and robust representation by modeling cross-scale spatial-temporal correlation. Specifically, CTL utilizes a CNN backbone and a key-points estimator to extract semantic local features from human body at multiple granularities as graph nodes. It explores a context-reinforced topology to construct multi-scale graphs by considering both global contextual information and physical connections of human body. Moreover, a 3D graph convolution and a cross-scale graph convolution are designed, which facilitate direct cross-spacetime and cross-scale information propagation for capturing hierarchical spatial-temporal dependencies and structural information. By jointly performing the two convolutions, CTL effectively mines comprehensive clues that are complementary with appearance information to enhance representational capacity. Extensive experiments on two video benchmarks have demonstrated the effectiveness of the proposed method and the state-of-the-art performance.

Unsupervised Domain Adaptive (UDA) person re-identification (ReID) aims at adapting the model trained on a labeled source-domain dataset to a target-domain dataset without any further annotations. Most successful UDA-ReID approaches combine clustering-based pseudo-label prediction with representation learning and perform the two steps in an alternating fashion. However, offline interaction between these two steps may allow noisy pseudo labels to substantially hinder the capability of the model. In this paper, we propose a Group-aware Label Transfer (GLT) algorithm, which enables the online interaction and mutual promotion of pseudo-label prediction and representation learning. Specifically, a label transfer algorithm simultaneously uses pseudo labels to train the data while refining the pseudo labels as an online clustering algorithm. It treats the online label refinery problem as an optimal transport problem, which explores the minimum cost for assigning M samples to N pseudo labels. More importantly, we introduce a group-aware strategy to assign implicit attribute group IDs to samples. The combination of the online label refining algorithm and the group-aware strategy can better correct the noisy pseudo label in an online fashion and narrow down the search space of the target identity. The effectiveness of the proposed GLT is demonstrated by the experimental results (Rank-1 accuracy) for Market1501→DukeMTMC (82.0%) and DukeMTMC→Market1501 (92.2%), remarkably closing the gap between unsupervised and supervised performance on person re-identification.1

Occluded person re-identification (ReID) aims to match person images with occlusion. It is fundamentally challenging because of the serious occlusion which aggravates the misalignment problem between images. At the cost of incorporating a pose estimator, many works introduce pose information to alleviate the misalignment in both training and testing. To achieve high accuracy while preserving low inference complexity, we propose a network named Pose-Guided Feature Learning with Knowledge Distillation (PGFL-KD), where the pose information is exploited to regularize the learning of semantics aligned features but is discarded in testing. PGFL-KD consists of a main branch (MB), and two pose-guided branches, e.g., a foreground-enhanced branch (FEB), and a body part semantics aligned branch (SAB). The FEB intends to emphasise the features of visible body parts while excluding the interference of obstructions and background (e.g., foreground feature alignment). The SAB encourages different channel groups to focus on different body parts to have body part semantics aligned representation. To get rid of the dependency on pose information when testing, we regularize the MB to learn the merits of the FEB and SAB through knowledge distillation and interaction-based training. Extensive experiments on occluded, partial, and holistic ReID tasks show the effectiveness of our proposed network.

Existing dominant approaches for cross-modal video-text retrieval task are to learn ajoint embedding space to measure the cross-modal similarity. However, these methods rarely explore long-range dependency inside video frames or textual words leading to insufficient textual and visual details. In this paper, we propose a stacked convolutional deep encoding network for video-text retrieval task, which considers to simultaneously encode long-range and short-range dependency in the videos and texts. Specifically, a multi-scale dilated convolutional (MSDC) block within our approach is able to encode short-range temporal cues between video frames or text words by adopting different scales of kernel size and dilation size of convolutional layer. A stacked structure is designed to expand the receptive fields by repeatedly adopting the MSD- C block, which further captures the long-range relations between these cues. Moreover, to obtain more robust textual representations, we fully utilize the powerful language model named Transformer in two stages: pretraining phrase and fine-tuning phrase. Extensive experiments on two different benchmark datasets (MSR-VTT, MSVD) show that our proposed method outperforms other state-of-the-art approaches.

Text-based person search aims to retrieve the pedestrian images that best match a given textual description from gallery images. Previous methods utilize the soft-attention mechanism to infer the semantic alignments between the regions of image and the corresponding words in sentence. However, these methods may fuse the irrelevant multi-modality features together which cause matching redundancy problem. In this work, we propose a novel hierarchical Gumbel attention network for text-based person search via Gumbel top-k re-parameterization algorithm. Specifically, it adaptively selects the strong semantically relevant image regions and words/phrases from images and texts for precise alignment and similarity calculation. This hard selection strategy is able to fuse the strong-relevant multi-modality features for alleviating the problem of matching redundancy. Meanwhile, a Gumbel top-k re-parameterization algorithm is designed as a low-variance, unbiased gradient estimator to handle the discreteness problem of hard attention mechanism by an end-to-end manner. Moreover, a hierarchical adaptive matching strategy is employed by the model from three different granularities, i.e., word-level, phrase-level, and sentence-level, towards fine-grained matching. Extensive experimental results demonstrate the state-of-the-art performance. Compared the existed best method, we achieve the 8.24% Rank-1 and 7.6% mAP relative improvements in the text-to-image retrieval task, and 5.58% Rank-1 and 6.3% mAP relative improvements in the image-to-text retrieval task on CUHK-PEDES dataset, respectively.

Abstraction reasoning is a long-standing challenge in artificial intelligence. Recent studies suggest that many of the deep architectures that have triumphed over other domains failed to work well in abstract reasoning. In this paper, we first illustrate that one of the main challenges in such a reasoning task is the presence of distracting features, which requires the learning algorithm to leverage counter-evidence and to reject any of false hypothesis in order to learn the true patterns. We later show that carefully designed learning trajectory over different categories of training data can effectively boost learning performance by mitigating the impacts of distracting features. Inspired this fact, we propose feature robust abstract reasoning (FRAR) model, which consists of a reinforcement learning based teacher network to determine the sequence of training and a student network for predictions. Experimental results demonstrated strong improvements over baseline algorithms and we are able to beat the state-of-the-art models by 18.7\% in RAVEN dataset and 13.3\% in the PGM dataset.

Depth estimation from monocular images is an ill-posed and inherently ambiguous problem. Recently, deep learning technique has been applied for monocular depth estimation seeking data-driven solutions. However, most existing methods focus on pursuing the minimization of average depth regression error at pixel level and neglect to encode the global layout of scene, resulting in layout-inconsistent depth map. This paper proposes a novel Layout-Aware Convolutional Neural Network (LA-Net) for accurate monocular depth estimation by simultaneously perceiving scene layout and local depth details. Specifically, a Spatial Layout Network (SL-Net) is proposed to learn a layout map representing the depth ordering between local patches. A Layout-Aware Depth Estimation Network (LDE-Net) is proposed to estimate pixel-level depth details using multi-scale layout maps as structural guidance, leading to layout-consistent depth map. A dense network module is used as the base network to learn effective visual details resorting to dense feed-forward connections. Moreover, we formulate an order-sensitive softmax loss to well constrain the ill-posed depth inferring problem. Extensive experiments on both indoor scene (NYUD-v2) and outdoor scene (Make3D) datasets have demonstrated that the proposed LA-Net outperforms the state-of-the-art methods and leads to faithful 3D projections.