Linjie Yang

Sijie Zhu, Linjie Yang, Chen Chen et al.

Visual Place Recognition (VPR) estimates the location of query images by matching them with images in a reference database. Conventional methods generally adopt aggregated CNN features for global retrieval and RANSAC-based geometric verification for reranking. However, RANSAC only employs geometric information but ignores other possible information that could be useful for reranking, e.g. local feature correlations, and attention values. In this paper, we propose a unified place recognition framework that handles both retrieval and reranking with a novel transformer model, named $R^{2}$Former. The proposed reranking module takes feature correlation, attention value, and xy coordinates into account, and learns to determine whether the image pair is from the same location. The whole pipeline is end-to-end trainable and the reranking module alone can also be adopted on other CNN or transformer backbones as a generic component. Remarkably, $R^{2}$Former significantly outperforms state-of-the-art methods on major VPR datasets with much less inference time and memory consumption. It also achieves the state-of-the-art on the hold-out MSLS challenge set and could serve as a simple yet strong solution for real-world large-scale applications. Experiments also show vision transformer tokens are comparable and sometimes better than CNN local features on local matching. The code is released at https://github.com/Jeff-Zilence/R2Former.

Cheng-En Wu, Yu Tian, Haichao Yu et al.

Vision-language models such as CLIP learn a generic text-image embedding from large-scale training data. A vision-language model can be adapted to a new classification task through few-shot prompt tuning. We find that such a prompt tuning process is highly robust to label noises. This intrigues us to study the key reasons contributing to the robustness of the prompt tuning paradigm. We conducted extensive experiments to explore this property and find the key factors are: 1) the fixed classname tokens provide a strong regularization to the optimization of the model, reducing gradients induced by the noisy samples; 2) the powerful pre-trained image-text embedding that is learned from diverse and generic web data provides strong prior knowledge for image classification. Further, we demonstrate that noisy zero-shot predictions from CLIP can be used to tune its own prompt, significantly enhancing prediction accuracy in the unsupervised setting. The code is available at https://github.com/CEWu/PTNL.

Taojiannan Yang, Linjie Yang, Xiaojie Jin et al.

Recent neural architecture search (NAS) works proposed training-free metrics to rank networks which largely reduced the search cost in NAS. In this paper, we revisit these training-free metrics and find that: (1) the number of parameters (\#Param), which is the most straightforward training-free metric, is overlooked in previous works but is surprisingly effective, (2) recent training-free metrics largely rely on the \#Param information to rank networks. Our experiments show that the performance of recent training-free metrics drops dramatically when the \#Param information is not available. Motivated by these observations, we argue that metrics less correlated with the \#Param are desired to provide additional information for NAS. We propose a light-weight training-based metric which has a weak correlation with the \#Param while achieving better performance than training-free metrics at a lower search cost. Specifically, on DARTS search space, our method completes searching directly on ImageNet in only 2.6 GPU hours and achieves a top-1/top-5 error rate of 24.1\%/7.1\%, which is competitive among state-of-the-art NAS methods. Codes are available at \url{https://github.com/taoyang1122/Revisit_TrainingFree_NAS}

Yuhan Shen, Linjie Yang, Longyin Wen et al.

Recent focus in video captioning has been on designing architectures that can consume both video and text modalities, and using large-scale video datasets with text transcripts for pre-training, such as HowTo100M. Though these approaches have achieved significant improvement, the audio modality is often ignored in video captioning. In this work, we present an audio-visual framework, which aims to fully exploit the potential of the audio modality for captioning. Instead of relying on text transcripts extracted via automatic speech recognition (ASR), we argue that learning with raw audio signals can be more beneficial, as audio has additional information including acoustic events, speaker identity, etc. Our contributions are twofold. First, we observed that the model overspecializes to the audio modality when pre-training with both video and audio modality, since the ground truth (i.e., text transcripts) can be solely predicted using audio. We proposed a Modality Balanced Pre-training (MBP) loss to mitigate this issue and significantly improve the performance on downstream tasks. Second, we slice and dice different design choices of the cross-modal module, which may become an information bottleneck and generate inferior results. We proposed new local-global fusion mechanisms to improve information exchange across audio and video. We demonstrate significant improvements by leveraging the audio modality on four datasets, and even outperform the state of the art on some metrics without relying on the text modality as the input.

Yan-Bo Lin, Yu Tian, Linjie Yang et al.

We present a framework for learning to generate background music from video inputs. Unlike existing works that rely on symbolic musical annotations, which are limited in quantity and diversity, our method leverages large-scale web videos accompanied by background music. This enables our model to learn to generate realistic and diverse music. To accomplish this goal, we develop a generative video-music Transformer with a novel semantic video-music alignment scheme. Our model uses a joint autoregressive and contrastive learning objective, which encourages the generation of music aligned with high-level video content. We also introduce a novel video-beat alignment scheme to match the generated music beats with the low-level motions in the video. Lastly, to capture fine-grained visual cues in a video needed for realistic background music generation, we introduce a new temporal video encoder architecture, allowing us to efficiently process videos consisting of many densely sampled frames. We train our framework on our newly curated DISCO-MV dataset, consisting of 2.2M video-music samples, which is orders of magnitude larger than any prior datasets used for video music generation. Our method outperforms existing approaches on the DISCO-MV and MusicCaps datasets according to various music generation evaluation metrics, including human evaluation. Results are available at https://genjib.github.io/project_page/VMAs/index.html

Haogeng Liu, Qihang Fan, Tingkai Liu et al.

This paper proposes Video-Teller, a video-language foundation model that leverages multi-modal fusion and fine-grained modality alignment to significantly enhance the video-to-text generation task. Video-Teller boosts the training efficiency by utilizing frozen pretrained vision and language modules. It capitalizes on the robust linguistic capabilities of large language models, enabling the generation of both concise and elaborate video descriptions. To effectively integrate visual and auditory information, Video-Teller builds upon the image-based BLIP-2 model and introduces a cascaded Q-Former which fuses information across frames and ASR texts. To better guide video summarization, we introduce a fine-grained modality alignment objective, where the cascaded Q-Former's output embedding is trained to align with the caption/summary embedding created by a pretrained text auto-encoder. Experimental results demonstrate the efficacy of our proposed video-language foundation model in accurately comprehending videos and generating coherent and precise language descriptions. It is worth noting that the fine-grained alignment enhances the model's capabilities (4% improvement of CIDEr score on MSR-VTT) with only 13% extra parameters in training and zero additional cost in inference.

Yiming Cui, Linjie Yang

Video object detection needs to solve feature degradation situations that rarely happen in the image domain. One solution is to use the temporal information and fuse the features from the neighboring frames. With Transformerbased object detectors getting a better performance on the image domain tasks, recent works began to extend those methods to video object detection. However, those existing Transformer-based video object detectors still follow the same pipeline as those used for classical object detectors, like enhancing the object feature representations by aggregation. In this work, we take a different perspective on video object detection. In detail, we improve the qualities of queries for the Transformer-based models by aggregation. To achieve this goal, we first propose a vanilla query aggregation module that weighted averages the queries according to the features of the neighboring frames. Then, we extend the vanilla module to a more practical version, which generates and aggregates queries according to the features of the input frames. Extensive experimental results validate the effectiveness of our proposed methods: On the challenging ImageNet VID benchmark, when integrated with our proposed modules, the current state-of-the-art Transformer-based object detectors can be improved by more than 2.4% on mAP and 4.2% on AP50.

Yiming Cui, Linjie Yang, Ding Liu

Object detection is a basic computer vision task to loccalize and categorize objects in a given image. Most state-of-the-art detection methods utilize a fixed number of proposals as an intermediate representation of object candidates, which is unable to adapt to different computational constraints during inference. In this paper, we propose a simple yet effective method which is adaptive to different computational resources by generating dynamic proposals for object detection. We first design a module to make a single query-based model to be able to inference with different numbers of proposals. Further, we extend it to a dynamic model to choose the number of proposals according to the input image, greatly reducing computational costs. Our method achieves significant speed-up across a wide range of detection models including two-stage and query-based models while obtaining similar or even better accuracy.

Yiming Cui, Linjie Yang, Haichao Yu

Transformer-based detection and segmentation methods use a list of learned detection queries to retrieve information from the transformer network and learn to predict the location and category of one specific object from each query. We empirically find that random convex combinations of the learned queries are still good for the corresponding models. We then propose to learn a convex combination with dynamic coefficients based on the high-level semantics of the image. The generated dynamic queries, named modulated queries, better capture the prior of object locations and categories in the different images. Equipped with our modulated queries, a wide range of DETR-based models achieve consistent and superior performance across multiple tasks including object detection, instance segmentation, panoptic segmentation, and video instance segmentation.

Xueqing Deng, Qi Fan, Xiaojie Jin et al.

Fine-tuning pre-trained transformer models, e.g., Swin Transformer, are successful in numerous downstream for dense prediction vision tasks. However, one major issue is the cost/storage of their huge amount of parameters, which becomes increasingly challenging to handle with the growing amount of vision tasks. In this paper, we propose an effective approach to alleviate the issue, namely selective feature adapter (SFA). It achieves state-of-the-art (SoTA) performance under any given budget of trainable parameters, and demonstrates comparable or better performance than fully fine-tuned models across various dense tasks. Specifically, SFA consists of external adapters and internal adapters which are sequentially operated over a transformer model. For external adapters, we properly select the places and amount of additional multilayer perception (MLP). For internal adapters, we transform a few task-important parameters inside the transformer, which are automatically discovered through a simple yet effective lottery ticket algorithm. Our experiments show that the dual adapter module, a.k.a SFA, is essential to achieve the best trade-off on dense vision tasks, such as segmentation, detection and depth-estimation, outperforming other adapters with a single module.

Henghui Ding, Lingyi Hong, Chang Liu et al.

Despite the promising performance of current video segmentation models on existing benchmarks, these models still struggle with complex scenes. In this paper, we introduce the 6th Large-scale Video Object Segmentation (LSVOS) challenge in conjunction with ECCV 2024 workshop. This year's challenge includes two tasks: Video Object Segmentation (VOS) and Referring Video Object Segmentation (RVOS). In this year, we replace the classic YouTube-VOS and YouTube-RVOS benchmark with latest datasets MOSE, LVOS, and MeViS to assess VOS under more challenging complex environments. This year's challenge attracted 129 registered teams from more than 20 institutes across over 8 countries. This report include the challenge and dataset introduction, and the methods used by top 7 teams in two tracks. More details can be found in our homepage https://lsvos.github.io/.

Haichao Yu, Yu Tian, Sateesh Kumar et al.

The quality of pre-training data plays a critical role in the performance of foundation models. Popular foundation models often design their own recipe for data filtering, which makes it hard to analyze and compare different data filtering approaches. DataComp is a new benchmark dedicated to evaluating different methods for data filtering. This paper describes our learning and solution when participating in the DataComp challenge. Our filtering strategy includes three stages: single-modality filtering, cross-modality filtering, and data distribution alignment. We integrate existing methods and propose new solutions, such as computing CLIP score on horizontally flipped images to mitigate the interference of scene text, using vision and language models to retrieve training samples for target downstream tasks, rebalancing the data distribution to improve the efficiency of allocating the computational budget, etc. We slice and dice our design choices, provide in-depth analysis, and discuss open questions. Our approach outperforms the best method from the DataComp paper by over 4% on the average performance of 38 tasks and by over 2% on ImageNet.

Khalil Mrini, Hanlin Lu, Linjie Yang et al.

Text-to-image generation has advanced rapidly, yet aligning complex textual prompts with generated visuals remains challenging, especially with intricate object relationships and fine-grained details. This paper introduces Fast Prompt Alignment (FPA), a prompt optimization framework that leverages a one-pass approach, enhancing text-to-image alignment efficiency without the iterative overhead typical of current methods like OPT2I. FPA uses large language models (LLMs) for single-iteration prompt paraphrasing, followed by fine-tuning or in-context learning with optimized prompts to enable real-time inference, reducing computational demands while preserving alignment fidelity. Extensive evaluations on the COCO Captions and PartiPrompts datasets demonstrate that FPA achieves competitive text-image alignment scores at a fraction of the processing time, as validated through both automated metrics (TIFA, VQA) and human evaluation. A human study with expert annotators further reveals a strong correlation between human alignment judgments and automated scores, underscoring the robustness of FPA's improvements. The proposed method showcases a scalable, efficient alternative to iterative prompt optimization, enabling broader applicability in real-time, high-demand settings. The codebase is provided to facilitate further research: https://github.com/tiktok/fast_prompt_alignment

Weizhi Wang, Yu Tian, Linjie Yang et al.

The reproduction of state-of-the-art multimodal LLM pre-training faces barriers at every stage of the pipeline, including high-quality data filtering, multimodal data mixture strategies, sequence packing techniques, and training frameworks. We introduce Open-Qwen2VL, a fully open-source 2B-parameter Multimodal Large Language Model pre-trained efficiently on 29M image-text pairs using only 220 A100-40G GPU hours. Our approach employs low-to-high dynamic image resolution and multimodal sequence packing to significantly enhance pre-training efficiency. The training dataset was carefully curated using both MLLM-based filtering techniques (e.g., MLM-Filter) and conventional CLIP-based filtering methods, substantially improving data quality and training efficiency. The Open-Qwen2VL pre-training is conducted on academic level 8xA100-40G GPUs at UCSB on 5B packed multimodal tokens, which is 0.36% of 1.4T multimodal pre-training tokens of Qwen2-VL. The final instruction-tuned Open-Qwen2VL outperforms partially-open state-of-the-art MLLM Qwen2-VL-2B on various multimodal benchmarks of MMBench, SEEDBench, MMstar, and MathVista, indicating the remarkable training efficiency of Open-Qwen2VL. We open-source all aspects of our work, including compute-efficient and data-efficient training details, data filtering methods, sequence packing scripts, pre-training data in WebDataset format, FSDP-based training codebase, and both base and instruction-tuned model checkpoints. We redefine "fully open" for multimodal LLMs as the complete release of: 1) the training codebase, 2) detailed data filtering techniques, and 3) all pre-training and supervised fine-tuning data used to develop the model.

Sucheng Ren, Xianhang Li, Haoqin Tu et al.

The vision community has started to build with the recently developed state space model, Mamba, as the new backbone for a range of tasks. This paper shows that Mamba's visual capability can be significantly enhanced through autoregressive pretraining, a direction not previously explored. Efficiency-wise, the autoregressive nature can well capitalize on the Mamba's unidirectional recurrent structure, enabling faster overall training speed compared to other training strategies like mask modeling. Performance-wise, autoregressive pretraining equips the Mamba architecture with markedly higher accuracy over its supervised-trained counterparts and, more importantly, successfully unlocks its scaling potential to large and even huge model sizes. For example, with autoregressive pretraining, a base-size Mamba attains 83.2\% ImageNet accuracy, outperforming its supervised counterpart by 2.0\%; our huge-size Mamba, the largest Vision Mamba to date, attains 85.0\% ImageNet accuracy (85.5\% when finetuned with $384\times384$ inputs), notably surpassing all other Mamba variants in vision. The code is available at \url{https://github.com/OliverRensu/ARM}.

Mingyu Ding, Xiaochen Lian, Linjie Yang et al.

High-resolution representations (HR) are essential for dense prediction tasks such as segmentation, detection, and pose estimation. Learning HR representations is typically ignored in previous Neural Architecture Search (NAS) methods that focus on image classification. This work proposes a novel NAS method, called HR-NAS, which is able to find efficient and accurate networks for different tasks, by effectively encoding multiscale contextual information while maintaining high-resolution representations. In HR-NAS, we renovate the NAS search space as well as its searching strategy. To better encode multiscale image contexts in the search space of HR-NAS, we first carefully design a lightweight transformer, whose computational complexity can be dynamically changed with respect to different objective functions and computation budgets. To maintain high-resolution representations of the learned networks, HR-NAS adopts a multi-branch architecture that provides convolutional encoding of multiple feature resolutions, inspired by HRNet. Last, we proposed an efficient fine-grained search strategy to train HR-NAS, which effectively explores the search space, and finds optimal architectures given various tasks and computation resources. HR-NAS is capable of achieving state-of-the-art trade-offs between performance and FLOPs for three dense prediction tasks and an image classification task, given only small computational budgets. For example, HR-NAS surpasses SqueezeNAS that is specially designed for semantic segmentation while improving efficiency by 45.9%. Code is available at https://github.com/dingmyu/HR-NAS

Xueyan Zou, Linjie Yang, Ding Liu et al.

Video inpainting aims to fill spatio-temporal "corrupted" regions with plausible content. To achieve this goal, it is necessary to find correspondences from neighbouring frames to faithfully hallucinate the unknown content. Current methods achieve this goal through attention, flow-based warping, or 3D temporal convolution. However, flow-based warping can create artifacts when optical flow is not accurate, while temporal convolution may suffer from spatial misalignment. We propose 'Progressive Temporal Feature Alignment Network', which progressively enriches features extracted from the current frame with the feature warped from neighbouring frames using optical flow. Our approach corrects the spatial misalignment in the temporal feature propagation stage, greatly improving visual quality and temporal consistency of the inpainted videos. Using the proposed architecture, we achieve state-of-the-art performance on the DAVIS and FVI datasets compared to existing deep learning approaches. Code is available at https://github.com/MaureenZOU/TSAM.

Mingyu Ding, Yuqi Huo, Haoyu Lu et al.

This work explores how to design a single neural network capable of adapting to multiple heterogeneous vision tasks, such as image segmentation, 3D detection, and video recognition. This goal is challenging because both network architecture search (NAS) spaces and methods in different tasks are inconsistent. We solve this challenge from both sides. We first introduce a unified design space for multiple tasks and build a multitask NAS benchmark (NAS-Bench-MR) on many widely used datasets, including ImageNet, Cityscapes, KITTI, and HMDB51. We further propose Network Coding Propagation (NCP), which back-propagates gradients of neural predictors to directly update architecture codes along the desired gradient directions to solve various tasks. In this way, optimal architecture configurations can be found by NCP in our large search space in seconds. Unlike prior arts of NAS that typically focus on a single task, NCP has several unique benefits. (1) NCP transforms architecture optimization from data-driven to architecture-driven, enabling joint search an architecture among multitasks with different data distributions. (2) NCP learns from network codes but not original data, enabling it to update the architecture efficiently across datasets. (3) In addition to our NAS-Bench-MR, NCP performs well on other NAS benchmarks, such as NAS-Bench-201. (4) Thorough studies of NCP on inter-, cross-, and intra-tasks highlight the importance of cross-task neural architecture design, i.e., multitask neural architectures and architecture transferring between different tasks. Code is available at https://github.com/dingmyu/NCP.

Daquan Zhou, Bingyi Kang, Xiaojie Jin et al.

Vision transformers (ViTs) have been successfully applied in image classification tasks recently. In this paper, we show that, unlike convolution neural networks (CNNs)that can be improved by stacking more convolutional layers, the performance of ViTs saturate fast when scaled to be deeper. More specifically, we empirically observe that such scaling difficulty is caused by the attention collapse issue: as the transformer goes deeper, the attention maps gradually become similar and even much the same after certain layers. In other words, the feature maps tend to be identical in the top layers of deep ViT models. This fact demonstrates that in deeper layers of ViTs, the self-attention mechanism fails to learn effective concepts for representation learning and hinders the model from getting expected performance gain. Based on above observation, we propose a simple yet effective method, named Re-attention, to re-generate the attention maps to increase their diversity at different layers with negligible computation and memory cost. The pro-posed method makes it feasible to train deeper ViT models with consistent performance improvements via minor modification to existing ViT models. Notably, when training a deep ViT model with 32 transformer blocks, the Top-1 classification accuracy can be improved by 1.6% on ImageNet. Code is publicly available at https://github.com/zhoudaquan/dvit_repo.

Yang Fu, Linjie Yang, Ding Liu et al.

Video instance segmentation is a complex task in which we need to detect, segment, and track each object for any given video. Previous approaches only utilize single-frame features for the detection, segmentation, and tracking of objects and they suffer in the video scenario due to several distinct challenges such as motion blur and drastic appearance change. To eliminate ambiguities introduced by only using single-frame features, we propose a novel comprehensive feature aggregation approach (CompFeat) to refine features at both frame-level and object-level with temporal and spatial context information. The aggregation process is carefully designed with a new attention mechanism which significantly increases the discriminative power of the learned features. We further improve the tracking capability of our model through a siamese design by incorporating both feature similarities and spatial similarities. Experiments conducted on the YouTube-VIS dataset validate the effectiveness of proposed CompFeat. Our code will be available at https://github.com/SHI-Labs/CompFeat-for-Video-Instance-Segmentation.

Yingwei Li, Xiaojie Jin, Jieru Mei et al.

Non-Local (NL) blocks have been widely studied in various vision tasks. However, it has been rarely explored to embed the NL blocks in mobile neural networks, mainly due to the following challenges: 1) NL blocks generally have heavy computation cost which makes it difficult to be applied in applications where computational resources are limited, and 2) it is an open problem to discover an optimal configuration to embed NL blocks into mobile neural networks. We propose AutoNL to overcome the above two obstacles. Firstly, we propose a Lightweight Non-Local (LightNL) block by squeezing the transformation operations and incorporating compact features. With the novel design choices, the proposed LightNL block is 400x computationally cheaper} than its conventional counterpart without sacrificing the performance. Secondly, by relaxing the structure of the LightNL block to be differentiable during training, we propose an efficient neural architecture search algorithm to learn an optimal configuration of LightNL blocks in an end-to-end manner. Notably, using only 32 GPU hours, the searched AutoNL model achieves 77.7% top-1 accuracy on ImageNet under a typical mobile setting (350M FLOPs), significantly outperforming previous mobile models including MobileNetV2 (+5.7%), FBNet (+2.8%) and MnasNet (+2.1%). Code and models are available at https://github.com/LiYingwei/AutoNL.

Jieru Mei, Yingwei Li, Xiaochen Lian et al.

Search space design is very critical to neural architecture search (NAS) algorithms. We propose a fine-grained search space comprised of atomic blocks, a minimal search unit that is much smaller than the ones used in recent NAS algorithms. This search space allows a mix of operations by composing different types of atomic blocks, while the search space in previous methods only allows homogeneous operations. Based on this search space, we propose a resource-aware architecture search framework which automatically assigns the computational resources (e.g., output channel numbers) for each operation by jointly considering the performance and the computational cost. In addition, to accelerate the search process, we propose a dynamic network shrinkage technique which prunes the atomic blocks with negligible influence on outputs on the fly. Instead of a search-and-retrain two-stage paradigm, our method simultaneously searches and trains the target architecture. Our method achieves state-of-the-art performance under several FLOPs configurations on ImageNet with a small searching cost. We open our entire codebase at: https://github.com/meijieru/AtomNAS.

Jiahui Yu, Linjie Yang, Ning Xu et al.

We present a simple and general method to train a single neural network executable at different widths (number of channels in a layer), permitting instant and adaptive accuracy-efficiency trade-offs at runtime. Instead of training individual networks with different width configurations, we train a shared network with switchable batch normalization. At runtime, the network can adjust its width on the fly according to on-device benchmarks and resource constraints, rather than downloading and offloading different models. Our trained networks, named slimmable neural networks, achieve similar (and in many cases better) ImageNet classification accuracy than individually trained models of MobileNet v1, MobileNet v2, ShuffleNet and ResNet-50 at different widths respectively. We also demonstrate better performance of slimmable models compared with individual ones across a wide range of applications including COCO bounding-box object detection, instance segmentation and person keypoint detection without tuning hyper-parameters. Lastly we visualize and discuss the learned features of slimmable networks. Code and models are available at: https://github.com/JiahuiYu/slimmable_networks

Zijie Li, Henry Li, Yichun Shi et al.

Diffusion models have gained tremendous success in text-to-image generation, yet still lag behind with visual understanding tasks, an area dominated by autoregressive vision-language models. We propose a large-scale and fully end-to-end diffusion model for multi-modal understanding and generation that significantly improves on existing diffusion-based multimodal models, and is the first of its kind to support the full suite of vision-language modeling capabilities. Inspired by the multimodal diffusion transformer (MM-DiT) and recent advances in discrete diffusion language modeling, we leverage a cross-modal maximum likelihood estimation framework that simultaneously trains the conditional likelihoods of both images and text jointly under a single loss function, which is back-propagated through both branches of the diffusion transformer. The resulting model is highly flexible and capable of a wide range of tasks including image generation, captioning, and visual question answering. Our model attained competitive performance compared to recent unified image understanding and generation models, demonstrating the potential of multimodal diffusion modeling as a promising alternative to autoregressive next-token prediction models.

Mingfei Han, Linjie Yang, Xiaojun Chang et al.

A short clip of video may contain progression of multiple events and an interesting story line. A human need to capture both the event in every shot and associate them together to understand the story behind it. In this work, we present a new multi-shot video understanding benchmark Shot2Story with detailed shot-level captions, comprehensive video summaries and question-answering pairs. To facilitate better semantic understanding of videos, we provide captions for both visual signals and human narrations. We design several distinct tasks including single-shot video captioning, multi-shot video summarization, and multi-shot video question answering. Preliminary experiments show some challenges to generate a long and comprehensive video summary for multi-shot videos. Nevertheless, the generated imperfect summaries can already achieve competitive performance on existing video understanding tasks such as video question-answering, promoting an under-explored setting of video understanding with detailed summaries.

Weizhi Wang, Khalil Mrini, Linjie Yang et al.

We propose a novel framework for filtering image-text data by leveraging fine-tuned Multimodal Language Models (MLMs). Our approach outperforms predominant filtering methods (e.g., CLIPScore) via integrating the recent advances in MLMs. We design four distinct yet complementary metrics to holistically measure the quality of image-text data. A new pipeline is established to construct high-quality instruction data for fine-tuning MLMs as data filters. Comparing with CLIPScore, our MLM filters produce more precise and comprehensive scores that directly improve the quality of filtered data and boost the performance of pre-trained models. We achieve significant improvements over CLIPScore on popular foundation models (i.e., CLIP and BLIP2) and various downstream tasks. Our MLM filter can generalize to different models and tasks, and be used as a drop-in replacement for CLIPScore. An additional ablation study is provided to verify our design choices for the MLM filter.

Lixue Gong, Xiaoxia Hou, Fanshi Li et al.

Rapid advancement of diffusion models has catalyzed remarkable progress in the field of image generation. However, prevalent models such as Flux, SD3.5 and Midjourney, still grapple with issues like model bias, limited text rendering capabilities, and insufficient understanding of Chinese cultural nuances. To address these limitations, we present Seedream 2.0, a native Chinese-English bilingual image generation foundation model that excels across diverse dimensions, which adeptly manages text prompt in both Chinese and English, supporting bilingual image generation and text rendering. We develop a powerful data system that facilitates knowledge integration, and a caption system that balances the accuracy and richness for image description. Particularly, Seedream is integrated with a self-developed bilingual large language model as a text encoder, allowing it to learn native knowledge directly from massive data. This enable it to generate high-fidelity images with accurate cultural nuances and aesthetic expressions described in either Chinese or English. Beside, Glyph-Aligned ByT5 is applied for flexible character-level text rendering, while a Scaled ROPE generalizes well to untrained resolutions. Multi-phase post-training optimizations, including SFT and RLHF iterations, further improve the overall capability. Through extensive experimentation, we demonstrate that Seedream 2.0 achieves state-of-the-art performance across multiple aspects, including prompt-following, aesthetics, text rendering, and structural correctness. Furthermore, Seedream 2.0 has been optimized through multiple RLHF iterations to closely align its output with human preferences, as revealed by its outstanding ELO score. In addition, it can be readily adapted to an instruction-based image editing model, such as SeedEdit, with strong editing capability that balances instruction-following and image consistency.

Wenda Chu, Bingliang Zhang, Jiaqi Han et al.

Autoregressive image modeling relies on visual tokenizers to compress images into compact latent representations. We design an end-to-end training pipeline that jointly optimizes reconstruction and generation, enabling direct supervision from generation results to the tokenizer. This contrasts with prior two-stage approaches that train tokenizers and generative models separately. We further investigate leveraging vision foundation models to improve 1D tokenizers for autoregressive modeling. Our autoregressive generative model achieves strong empirical results, including a state-of-the-art FID score of 1.48 without guidance on ImageNet 256x256 generation.

Xueqing Deng, Qihang Yu, Ali Athar et al.

This paper introduces the COCONut-PanCap dataset, created to enhance panoptic segmentation and grounded image captioning. Building upon the COCO dataset with advanced COCONut panoptic masks, this dataset aims to overcome limitations in existing image-text datasets that often lack detailed, scene-comprehensive descriptions. The COCONut-PanCap dataset incorporates fine-grained, region-level captions grounded in panoptic segmentation masks, ensuring consistency and improving the detail of generated captions. Through human-edited, densely annotated descriptions, COCONut-PanCap supports improved training of vision-language models (VLMs) for image understanding and generative models for text-to-image tasks. Experimental results demonstrate that COCONut-PanCap significantly boosts performance across understanding and generation tasks, offering complementary benefits to large-scale datasets. This dataset sets a new benchmark for evaluating models on joint panoptic segmentation and grounded captioning tasks, addressing the need for high-quality, detailed image-text annotations in multi-modal learning.

Mingfei Han, Linjie Yang, Xiaojie Jin et al.

The creation of new datasets often presents new challenges for video recognition and can inspire novel ideas while addressing these challenges. While existing datasets mainly comprise landscape mode videos, our paper seeks to introduce portrait mode videos to the research community and highlight the unique challenges associated with this video format. With the growing popularity of smartphones and social media applications, recognizing portrait mode videos is becoming increasingly important. To this end, we have developed the first dataset dedicated to portrait mode video recognition, namely PortraitMode-400. The taxonomy of PortraitMode-400 was constructed in a data-driven manner, comprising 400 fine-grained categories, and rigorous quality assurance was implemented to ensure the accuracy of human annotations. In addition to the new dataset, we conducted a comprehensive analysis of the impact of video format (portrait mode versus landscape mode) on recognition accuracy and spatial bias due to the different formats. Furthermore, we designed extensive experiments to explore key aspects of portrait mode video recognition, including the choice of data augmentation, evaluation procedure, the importance of temporal information, and the role of audio modality. Building on the insights from our experimental results and the introduction of PortraitMode-400, our paper aims to inspire further research efforts in this emerging research area.

Zhiheng Liu, Xueqing Deng, Shoufa Chen et al.

Generative video modeling has made significant strides, yet ensuring structural and temporal consistency over long sequences remains a challenge. Current methods predominantly rely on RGB signals, leading to accumulated errors in object structure and motion over extended durations. To address these issues, we introduce WorldWeaver, a robust framework for long video generation that jointly models RGB frames and perceptual conditions within a unified long-horizon modeling scheme. Our training framework offers three key advantages. First, by jointly predicting perceptual conditions and color information from a unified representation, it significantly enhances temporal consistency and motion dynamics. Second, by leveraging depth cues, which we observe to be more resistant to drift than RGB, we construct a memory bank that preserves clearer contextual information, improving quality in long-horizon video generation. Third, we employ segmented noise scheduling for training prediction groups, which further mitigates drift and reduces computational cost. Extensive experiments on both diffusion- and rectified flow-based models demonstrate the effectiveness of WorldWeaver in reducing temporal drift and improving the fidelity of generated videos.

Chang Liu, Henghui Ding, Kaining Ying et al.

This report presents an overview of the 7th Large-scale Video Object Segmentation (LSVOS) Challenge held in conjunction with ICCV 2025. Besides the two traditional tracks of LSVOS that jointly target robustness in realistic video scenarios: Classic VOS (VOS), and Referring VOS (RVOS), the 2025 edition features a newly introduced track, Complex VOS (MOSEv2). Building upon prior insights, MOSEv2 substantially increases difficulty, introducing more challenging but realistic scenarios including denser small objects, frequent disappear/reappear events, severe occlusions, adverse weather and lighting, etc., pushing long-term consistency and generalization beyond curated benchmarks. The challenge retains standard ${J}$, $F$, and ${J\&F}$ metrics for VOS and RVOS, while MOSEv2 adopts ${J\&\dot{F}}$ as the primary ranking metric to better evaluate objects across scales and disappearance cases. We summarize datasets and protocols, highlight top-performing solutions, and distill emerging trends, such as the growing role of LLM/MLLM components and memory-aware propagation, aiming to chart future directions for resilient, language-aware video segmentation in the wild.

Mingfei Han, Haihong Hao, Jinxing Zhou et al.

Vision-language models often hallucinate details, generating non-existent objects or inaccurate attributes that compromise output reliability. Existing methods typically address these issues via extensive human annotations or external supervision from more powerful models. In this work, we present a novel framework that leverages the model's self-consistency between long responses and short answers to generate preference pairs for training. We observe that short binary questions tend to yield highly reliable responses, which can be used to query the target model to evaluate and rank its generated responses. Specifically, we design a self-reflection pipeline where detailed model responses are compared against concise binary answers, and inconsistency signals are utilized to automatically curate high-quality training data without human annotations or external model-based supervision. By relying solely on self-consistency rather than external supervision, our method offers a scalable and efficient solution that effectively reduces hallucinations using unlabeled data. Extensive experiments on multiple benchmarks, i.e., AMBER, MultiObject-Hal (ROPE), Object HalBench, and MMHal-Bench, demonstrate significant improvements in factual grounding and reliability. Moreover, our approach maintains robust instruction-following ability, as evidenced by enhanced performance on LLaVA-Bench and MMBench.

Shanchuan Lin, Linjie Yang, Imran Saleemi et al.

We introduce a robust, real-time, high-resolution human video matting method that achieves new state-of-the-art performance. Our method is much lighter than previous approaches and can process 4K at 76 FPS and HD at 104 FPS on an Nvidia GTX 1080Ti GPU. Unlike most existing methods that perform video matting frame-by-frame as independent images, our method uses a recurrent architecture to exploit temporal information in videos and achieves significant improvements in temporal coherence and matting quality. Furthermore, we propose a novel training strategy that enforces our network on both matting and segmentation objectives. This significantly improves our model's robustness. Our method does not require any auxiliary inputs such as a trimap or a pre-captured background image, so it can be widely applied to existing human matting applications.

Haichao Yu, Linjie Yang, Humphrey Shi

Post-training quantization methods use a set of calibration data to compute quantization ranges for network parameters and activations. The calibration data usually comes from the training dataset which could be inaccessible due to sensitivity of the data. In this work, we want to study such a problem: can we use out-of-domain data to calibrate the trained networks without knowledge of the original dataset? Specifically, we go beyond the domain of natural images to include drastically different domains such as X-ray images, satellite images and ultrasound images. We find cross-domain calibration leads to surprisingly stable performance of quantized models on 10 tasks in different image domains with 13 different calibration datasets. We also find that the performance of quantized models is correlated with the similarity of the Gram matrices between the source and calibration domains, which can be used as a criterion to choose calibration set for better performance. We believe our research opens the door to borrow cross-domain knowledge for network quantization and compression.

Daquan Zhou, Xiaojie Jin, Xiaochen Lian et al.

Current neural architecture search (NAS) algorithms still require expert knowledge and effort to design a search space for network construction. In this paper, we consider automating the search space design to minimize human interference, which however faces two challenges: the explosive complexity of the exploration space and the expensive computation cost to evaluate the quality of different search spaces. To solve them, we propose a novel differentiable evolutionary framework named AutoSpace, which evolves the search space to an optimal one with following novel techniques: a differentiable fitness scoring function to efficiently evaluate the performance of cells and a reference architecture to speedup the evolution procedure and avoid falling into sub-optimal solutions. The framework is generic and compatible with additional computational constraints, making it feasible to learn specialized search spaces that fit different computational budgets. With the learned search space, the performance of recent NAS algorithms can be improved significantly compared with using previously manually designed spaces. Remarkably, the models generated from the new search space achieve 77.8% top-1 accuracy on ImageNet under the mobile setting (MAdds < 500M), out-performing previous SOTA EfficientNet-B0 by 0.7%. All codes will be made public.

Linjie Yang, Qing Jin

Model quantization helps to reduce model size and latency of deep neural networks. Mixed precision quantization is favorable with customized hardwares supporting arithmetic operations at multiple bit-widths to achieve maximum efficiency. We propose a novel learning-based algorithm to derive mixed precision models end-to-end under target computation constraints and model sizes. During the optimization, the bit-width of each layer / kernel in the model is at a fractional status of two consecutive bit-widths which can be adjusted gradually. With a differentiable regularization term, the resource constraints can be met during the quantization-aware training which results in an optimized mixed precision model. Further, our method can be naturally combined with channel pruning for better computation cost allocation. Our final models achieve comparable or better performance than previous quantization methods with mixed precision on MobilenetV1/V2, ResNet18 under different resource constraints on ImageNet dataset.

Qing Jin, Linjie Yang, Zhenyu Liao

Quantization reduces computation costs of neural networks but suffers from performance degeneration. Is this accuracy drop due to the reduced capacity, or inefficient training during the quantization procedure? After looking into the gradient propagation process of neural networks by viewing the weights and intermediate activations as random variables, we discover two critical rules for efficient training. Recent quantization approaches violates the two rules and results in degenerated convergence. To deal with this problem, we propose a simple yet effective technique, named scale-adjusted training (SAT), to comply with the discovered rules and facilitates efficient training. We also analyze the quantization error introduced in calculating the gradient in the popular parameterized clipping activation (PACT) technique. Through SAT together with gradient-calibrated PACT, quantized models obtain comparable or even better performance than their full-precision counterparts, achieving state-of-the-art accuracy with consistent improvement over previous quantization methods on a wide spectrum of models including MobileNet-V1/V2 and PreResNet-50.

Qing Jin, Linjie Yang, Zhenyu Liao

Deep neural networks with adaptive configurations have gained increasing attention due to the instant and flexible deployment of these models on platforms with different resource budgets. In this paper, we investigate a novel option to achieve this goal by enabling adaptive bit-widths of weights and activations in the model. We first examine the benefits and challenges of training quantized model with adaptive bit-widths, and then experiment with several approaches including direct adaptation, progressive training and joint training. We discover that joint training is able to produce comparable performance on the adaptive model as individual models. We further propose a new technique named Switchable Clipping Level (S-CL) to further improve quantized models at the lowest bit-width. With our proposed techniques applied on a bunch of models including MobileNet-V1/V2 and ResNet-50, we demonstrate that bit-width of weights and activations is a new option for adaptively executable deep neural networks, offering a distinct opportunity for improved accuracy-efficiency trade-off as well as instant adaptation according to the platform constraints in real-world applications.

Zhengyuan Yang, Yuncheng Li, Linjie Yang et al.

Human body part segmentation refers to the task of predicting the semantic segmentation mask for each body part. Fully supervised body part segmentation methods achieve good performances but require an enormous amount of effort to annotate part masks for training. In contrast to high annotation costs needed for a limited number of part mask annotations, a large number of weak labels such as poses and full body masks already exist and contain relevant information. Motivated by the possibility of using existing weak labels, we propose the first weakly supervised body part segmentation framework. The core idea is first converting the sparse weak labels such as keypoints to the initial estimate of body part masks, and then iteratively refine the part mask predictions. We name the initial part masks estimated from poses the "part priors." With sufficient extra weak labels, our weakly supervised framework achieves a comparable performance (62.0% mIoU) to the fully supervised method (63.6% mIoU) on the Pascal-Person-Part dataset. Furthermore, in the extended semi-supervised setting, the proposed framework outperforms the state-of-art methods. Moreover, we extend our proposed framework to other keypoint-supervised part segmentation tasks such as face parsing.

Linjie Yang, Yuchen Fan, Ning Xu

In this paper we present a new computer vision task, named video instance segmentation. The goal of this new task is simultaneous detection, segmentation and tracking of instances in videos. In words, it is the first time that the image instance segmentation problem is extended to the video domain. To facilitate research on this new task, we propose a large-scale benchmark called YouTube-VIS, which consists of 2883 high-resolution YouTube videos, a 40-category label set and 131k high-quality instance masks. In addition, we propose a novel algorithm called MaskTrack R-CNN for this task. Our new method introduces a new tracking branch to Mask R-CNN to jointly perform the detection, segmentation and tracking tasks simultaneously. Finally, we evaluate the proposed method and several strong baselines on our new dataset. Experimental results clearly demonstrate the advantages of the proposed algorithm and reveal insight for future improvement. We believe the video instance segmentation task will motivate the community along the line of research for video understanding.

Ruotian Luo, Ning Zhang, Bohyung Han et al.

We present a novel problem setting in zero-shot learning, zero-shot object recognition and detection in the context. Contrary to the traditional zero-shot learning methods, which simply infers unseen categories by transferring knowledge from the objects belonging to semantically similar seen categories, we aim to understand the identity of the novel objects in an image surrounded by the known objects using the inter-object relation prior. Specifically, we leverage the visual context and the geometric relationships between all pairs of objects in a single image, and capture the information useful to infer unseen categories. We integrate our context-aware zero-shot learning framework into the traditional zero-shot learning techniques seamlessly using a Conditional Random Field (CRF). The proposed algorithm is evaluated on both zero-shot region classification and zero-shot detection tasks. The results on Visual Genome (VG) dataset show that our model significantly boosts performance with the additional visual context compared to traditional methods.

Jonghwan Mun, Linjie Yang, Zhou Ren et al.

Dense video captioning is an extremely challenging task since accurate and coherent description of events in a video requires holistic understanding of video contents as well as contextual reasoning of individual events. Most existing approaches handle this problem by first detecting event proposals from a video and then captioning on a subset of the proposals. As a result, the generated sentences are prone to be redundant or inconsistent since they fail to consider temporal dependency between events. To tackle this challenge, we propose a novel dense video captioning framework, which models temporal dependency across events in a video explicitly and leverages visual and linguistic context from prior events for coherent storytelling. This objective is achieved by 1) integrating an event sequence generation network to select a sequence of event proposals adaptively, and 2) feeding the sequence of event proposals to our sequential video captioning network, which is trained by reinforcement learning with two-level rewards at both event and episode levels for better context modeling. The proposed technique achieves outstanding performances on ActivityNet Captions dataset in most metrics.

Ning Xu, Linjie Yang, Yuchen Fan et al.

Learning long-term spatial-temporal features are critical for many video analysis tasks. However, existing video segmentation methods predominantly rely on static image segmentation techniques, and methods capturing temporal dependency for segmentation have to depend on pretrained optical flow models, leading to suboptimal solutions for the problem. End-to-end sequential learning to explore spatialtemporal features for video segmentation is largely limited by the scale of available video segmentation datasets, i.e., even the largest video segmentation dataset only contains 90 short video clips. To solve this problem, we build a new large-scale video object segmentation dataset called YouTube Video Object Segmentation dataset (YouTube-VOS). Our dataset contains 4,453 YouTube video clips and 94 object categories. This is by far the largest video object segmentation dataset to our knowledge and has been released at http://youtube-vos.org. We further evaluate several existing state-of-the-art video object segmentation algorithms on this dataset which aims to establish baselines for the development of new algorithms in the future.

Ning Xu, Linjie Yang, Yuchen Fan et al.

Learning long-term spatial-temporal features are critical for many video analysis tasks. However, existing video segmentation methods predominantly rely on static image segmentation techniques, and methods capturing temporal dependency for segmentation have to depend on pretrained optical flow models, leading to suboptimal solutions for the problem. End-to-end sequential learning to explore spatial-temporal features for video segmentation is largely limited by the scale of available video segmentation datasets, i.e., even the largest video segmentation dataset only contains 90 short video clips. To solve this problem, we build a new large-scale video object segmentation dataset called YouTube Video Object Segmentation dataset (YouTube-VOS). Our dataset contains 3,252 YouTube video clips and 78 categories including common objects and human activities. This is by far the largest video object segmentation dataset to our knowledge and we have released it at https://youtube-vos.org. Based on this dataset, we propose a novel sequence-to-sequence network to fully exploit long-term spatial-temporal information in videos for segmentation. We demonstrate that our method is able to achieve the best results on our YouTube-VOS test set and comparable results on DAVIS 2016 compared to the current state-of-the-art methods. Experiments show that the large scale dataset is indeed a key factor to the success of our model.

Linjie Yang, Yanran Wang, Xuehan Xiong et al.

Video object segmentation targets at segmenting a specific object throughout a video sequence, given only an annotated first frame. Recent deep learning based approaches find it effective by fine-tuning a general-purpose segmentation model on the annotated frame using hundreds of iterations of gradient descent. Despite the high accuracy these methods achieve, the fine-tuning process is inefficient and fail to meet the requirements of real world applications. We propose a novel approach that uses a single forward pass to adapt the segmentation model to the appearance of a specific object. Specifically, a second meta neural network named modulator is learned to manipulate the intermediate layers of the segmentation network given limited visual and spatial information of the target object. The experiments show that our approach is 70times faster than fine-tuning approaches while achieving similar accuracy.

Linjie Yang, Kevin Tang, Jianchao Yang et al.

Dense captioning is a newly emerging computer vision topic for understanding images with dense language descriptions. The goal is to densely detect visual concepts (e.g., objects, object parts, and interactions between them) from images, labeling each with a short descriptive phrase. We identify two key challenges of dense captioning that need to be properly addressed when tackling the problem. First, dense visual concept annotations in each image are associated with highly overlapping target regions, making accurate localization of each visual concept challenging. Second, the large amount of visual concepts makes it hard to recognize each of them by appearance alone. We propose a new model pipeline based on two novel ideas, joint inference and context fusion, to alleviate these two challenges. We design our model architecture in a methodical manner and thoroughly evaluate the variations in architecture. Our final model, compact and efficient, achieves state-of-the-art accuracy on Visual Genome for dense captioning with a relative gain of 73\% compared to the previous best algorithm. Qualitative experiments also reveal the semantic capabilities of our model in dense captioning.

Linjie Yang, Ping Luo, Chen Change Loy et al.

Updated on 24/09/2015: This update provides preliminary experiment results for fine-grained classification on the surveillance data of CompCars. The train/test splits are provided in the updated dataset. See details in Section 6.