Zhaoqing Wang

Shaokun Zhang, Xiaobo Xia, Zhaoqing Wang et al. · tsinghua

In-context learning is a promising paradigm that utilizes in-context examples as prompts for the predictions of large language models. These prompts are crucial for achieving strong performance. However, since the prompts need to be sampled from a large volume of annotated examples, finding the right prompt may result in high annotation costs. To address this challenge, this paper introduces an influence-driven selective annotation method that aims to minimize annotation costs while improving the quality of in-context examples. The essence of our method is to select a pivotal subset from a large-scale unlabeled data pool to annotate for the subsequent sampling of prompts. Specifically, a directed graph is first constructed to represent unlabeled data. Afterward, the influence of candidate unlabeled subsets is quantified with a diffusion process. A simple yet effective greedy algorithm for unlabeled data selection is lastly introduced. It iteratively selects the data if it provides a maximum marginal gain with respect to quantified influence. Compared with previous efforts on selective annotations, our influence-driven method works in an end-to-end manner, avoids an intractable explicit balance between data diversity and representativeness, and enjoys theoretical support. Experiments confirm the superiority of the proposed method on various benchmarks, achieving better performance under lower time consumption during subset selection. The project page is available at https://skzhang1.github.io/IDEAL/.

Xiaowei Chi, Jiaming Liu, Ming Lu et al.

Bird's-Eye-View (BEV) 3D Object Detection is a crucial multi-view technique for autonomous driving systems. Recently, plenty of works are proposed, following a similar paradigm consisting of three essential components, i.e., camera feature extraction, BEV feature construction, and task heads. Among the three components, BEV feature construction is BEV-specific compared with 2D tasks. Existing methods aggregate the multi-view camera features to the flattened grid in order to construct the BEV feature. However, flattening the BEV space along the height dimension fails to emphasize the informative features of different heights. For example, the barrier is located at a low height while the truck is located at a high height. In this paper, we propose a novel method named BEV Slice Attention Network (BEV-SAN) for exploiting the intrinsic characteristics of different heights. Instead of flattening the BEV space, we first sample along the height dimension to build the global and local BEV slices. Then, the features of BEV slices are aggregated from the camera features and merged by the attention mechanism. Finally, we fuse the merged local and global BEV features by a transformer to generate the final feature map for task heads. The purpose of local BEV slices is to emphasize informative heights. In order to find them, we further propose a LiDAR-guided sampling strategy to leverage the statistical distribution of LiDAR to determine the heights of local slices. Compared with uniform sampling, LiDAR-guided sampling can determine more informative heights. We conduct detailed experiments to demonstrate the effectiveness of BEV-SAN. Code will be released.

Yingbin Bai, Erkun Yang, Zhaoqing Wang et al.

Most recent self-supervised learning methods learn visual representation by contrasting different augmented views of images. Compared with supervised learning, more aggressive augmentations have been introduced to further improve the diversity of training pairs. However, aggressive augmentations may distort images' structures leading to a severe semantic shift problem that augmented views of the same image may not share the same semantics, thus degrading the transfer performance. To address this problem, we propose a new SSL paradigm, which counteracts the impact of semantic shift by balancing the role of weak and aggressively augmented pairs. Specifically, semantically inconsistent pairs are of minority and we treat them as noisy pairs. Note that deep neural networks (DNNs) have a crucial memorization effect that DNNs tend to first memorize clean (majority) examples before overfitting to noisy (minority) examples. Therefore, we set a relatively large weight for aggressively augmented data pairs at the early learning stage. With the training going on, the model begins to overfit noisy pairs. Accordingly, we gradually reduce the weights of aggressively augmented pairs. In doing so, our method can better embrace the aggressive augmentations and neutralize the semantic shift problem. Experiments show that our model achieves 73.1% top-1 accuracy on ImageNet-1K with ResNet-50 for 200 epochs, which is a 2.5% improvement over BYOL. Moreover, experiments also demonstrate that the learned representations can transfer well for various downstream tasks.

Sensen Gao, Zhaoqing Wang, Qihang Cao et al.

Existing diffusion-based 3D scene generation methods primarily operate in 2D image/video latent spaces, which makes maintaining cross-view appearance and geometric consistency inherently challenging. To bridge this gap, we present OneWorld, a framework that performs diffusion directly within a coherent 3D representation space. Central to our approach is the 3D Unified Representation Autoencoder (3D-URAE); it leverages pretrained 3D foundation models and augments their geometry-centric nature by injecting appearance and distilling semantics into a unified 3D latent space. Furthermore, we introduce token-level Cross-View-Correspondence (CVC) consistency loss to explicitly enforce structural alignment across views, and propose Manifold-Drift Forcing (MDF) to mitigate train-inference exposure bias and shape a robust 3D manifold by mixing drifted and original representations. Comprehensive experiments demonstrate that OneWorld generates high-quality 3D scenes with superior cross-view consistency compared to state-of-the-art 2D-based methods. Our code will be available at https://github.com/SensenGao/OneWorld.

Runnan Chen, Xiangyu Sun, Zhaoqing Wang et al.

Open-vocabulary scene understanding using 3D Gaussian (3DGS) representations has garnered considerable attention. However, existing methods mostly lift knowledge from large 2D vision models into 3DGS on a scene-by-scene basis, restricting the capabilities of open-vocabulary querying within their training scenes so that lacking the generalizability to novel scenes. In this work, we propose \textbf{OVGaussian}, a generalizable \textbf{O}pen-\textbf{V}ocabulary 3D semantic segmentation framework based on the 3D \textbf{Gaussian} representation. We first construct a large-scale 3D scene dataset based on 3DGS, dubbed \textbf{SegGaussian}, which provides detailed semantic and instance annotations for both Gaussian points and multi-view images. To promote semantic generalization across scenes, we introduce Generalizable Semantic Rasterization (GSR), which leverages a 3D neural network to learn and predict the semantic property for each 3D Gaussian point, where the semantic property can be rendered as multi-view consistent 2D semantic maps. In the next, we propose a Cross-modal Consistency Learning (CCL) framework that utilizes open-vocabulary annotations of 2D images and 3D Gaussians within SegGaussian to train the 3D neural network capable of open-vocabulary semantic segmentation across Gaussian-based 3D scenes. Experimental results demonstrate that OVGaussian significantly outperforms baseline methods, exhibiting robust cross-scene, cross-domain, and novel-view generalization capabilities. Code and the SegGaussian dataset will be released. (https://github.com/runnanchen/OVGaussian).

Runnan Chen, Zhaoqing Wang, Jiepeng Wang et al.

Understanding geometric, semantic, and instance information in 3D scenes from sequential video data is essential for applications in robotics and augmented reality. However, existing Simultaneous Localization and Mapping (SLAM) methods generally focus on either geometric or semantic reconstruction. In this paper, we introduce PanoSLAM, the first SLAM system to integrate geometric reconstruction, 3D semantic segmentation, and 3D instance segmentation within a unified framework. Our approach builds upon 3D Gaussian Splatting, modified with several critical components to enable efficient rendering of depth, color, semantic, and instance information from arbitrary viewpoints. To achieve panoptic 3D scene reconstruction from sequential RGB-D videos, we propose an online Spatial-Temporal Lifting (STL) module that transfers 2D panoptic predictions from vision models into 3D Gaussian representations. This STL module addresses the challenges of label noise and inconsistencies in 2D predictions by refining the pseudo labels across multi-view inputs, creating a coherent 3D representation that enhances segmentation accuracy. Our experiments show that PanoSLAM outperforms recent semantic SLAM methods in both mapping and tracking accuracy. For the first time, it achieves panoptic 3D reconstruction of open-world environments directly from the RGB-D video. (https://github.com/runnanchen/PanoSLAM)

Yongli Xiang, Ziming Hong, Zhaoqing Wang et al.

Text-to-Image (T2I) diffusion models have demonstrated significant advancements in generating high-quality images, while raising potential safety concerns regarding harmful content generation. Safety-guidance-based methods have been proposed to mitigate harmful outputs by steering generation away from harmful zones, where the zones are averaged across multiple harmful categories based on predefined keywords. However, these approaches fail to capture the complex interplay among different harm categories, leading to "harmful conflicts" where mitigating one type of harm may inadvertently amplify another, thus increasing overall harmful rate. To address this issue, we propose Conflict-aware Adaptive Safety Guidance (CASG), a training-free framework that dynamically identifies and applies the category-aligned safety direction during generation. CASG is composed of two components: (i) Conflict-aware Category Identification (CaCI), which identifies the harmful category most aligned with the model's evolving generative state, and (ii) Conflict-resolving Guidance Application (CrGA), which applies safety steering solely along the identified category to avoid multi-category interference. CASG can be applied to both latent-space and text-space safeguards. Experiments on T2I safety benchmarks demonstrate CASG's state-of-the-art performance, reducing the harmful rate by up to 15.4% compared to existing methods.

Jiaming Liu, Ming Lu, Kaixin Chen et al.

Internet video delivery has undergone a tremendous explosion of growth over the past few years. However, the quality of video delivery system greatly depends on the Internet bandwidth. Deep Neural Networks (DNNs) are utilized to improve the quality of video delivery recently. These methods divide a video into chunks, and stream LR video chunks and corresponding content-aware models to the client. The client runs the inference of models to super-resolve the LR chunks. Consequently, a large number of models are streamed in order to deliver a video. In this paper, we first carefully study the relation between models of different chunks, then we tactfully design a joint training framework along with the Content-aware Feature Modulation (CaFM) layer to compress these models for neural video delivery. {\bf With our method, each video chunk only requires less than $1\% $ of original parameters to be streamed, achieving even better SR performance.} We conduct extensive experiments across various SR backbones, video time length, and scaling factors to demonstrate the advantages of our method. Besides, our method can be also viewed as a new approach of video coding. Our primary experiments achieve better video quality compared with the commercial H.264 and H.265 standard under the same storage cost, showing the great potential of the proposed method. Code is available at:\url{https://github.com/Neural-video-delivery/CaFM-Pytorch-ICCV2021}

Zhaoqing Wang, Xiaobo Xia, Zhuolin Bie et al.

Recent advances in video diffusion models have remarkably improved camera-controlled video generation, but most methods rely solely on supervised fine-tuning (SFT), leaving online reinforcement learning (RL) post-training largely underexplored. In this work, we introduce an online RL post-training framework that optimizes a pretrained video generator for precise camera control. To make RL effective in this setting, we design a verifiable geometry reward that delivers dense segment-level feedback to guide model optimization. Specifically, we estimate the 3D camera trajectories for both generated and reference videos, divide each trajectory into short segments, and compute segment-wise relative poses. The reward function then compares each generated-reference segment pair and assigns an alignment score as the reward signal, which helps alleviate reward sparsity and improve optimization efficiency. Moreover, we construct a comprehensive dataset featuring diverse large-amplitude camera motions and scenes with varied subject dynamics. Extensive experiments show that our online RL post-training clearly outperforms SFT baselines across multiple aspects, including camera-control accuracy, geometric consistency, and visual quality, demonstrating its superiority in advancing camera-controlled video generation.

Zhaoqing Wang, Xiaobo Xia, Runnan Chen et al.

This paper presents the Large Vision Diffusion Transformer (LaVin-DiT), a scalable and unified foundation model designed to tackle over 20 computer vision tasks in a generative framework. Unlike existing large vision models directly adapted from natural language processing architectures, which rely on less efficient autoregressive techniques and disrupt spatial relationships essential for vision data, LaVin-DiT introduces key innovations to optimize generative performance for vision tasks. First, to address the high dimensionality of visual data, we incorporate a spatial-temporal variational autoencoder that encodes data into a continuous latent space. Second, for generative modeling, we develop a joint diffusion transformer that progressively produces vision outputs. Third, for unified multi-task training, in-context learning is implemented. Input-target pairs serve as task context, which guides the diffusion transformer to align outputs with specific tasks within the latent space. During inference, a task-specific context set and test data as queries allow LaVin-DiT to generalize across tasks without fine-tuning. Trained on extensive vision datasets, the model is scaled from 0.1B to 3.4B parameters, demonstrating substantial scalability and state-of-the-art performance across diverse vision tasks. This work introduces a novel pathway for large vision foundation models, underscoring the promising potential of diffusion transformers. The code and models are available.

Zhaoqing Wang, Xiaobo Xia, Ziye Chen et al.

Current state-of-the-art open-vocabulary segmentation methods typically rely on image-mask-text triplet annotations for supervision. However, acquiring such detailed annotations is labour-intensive and poses scalability challenges in complex real-world scenarios. While existing weakly-supervised approaches leverage image-text pairs to reduce the expansive annotation cost, the lack of mask supervision makes it difficult for the model to locate multiple instances and accurately group pixels with similar semantics, significantly hampering versatility and performance. In this paper, we introduce Unpair-Seg, a novel weakly-supervised open-vocabulary segmentation framework that learns from unpaired image-mask and image-text pairs, which can be independently and efficiently collected. Unpair-Seg initially predicts a set of binary masks and generates pseudo labels by identifying confident pairs of masks and text entities. We then train a feature adapter to align region embeddings with text embeddings based on these pseudo labels, achieving open-vocabulary segmentation. However, the inherent noise in the mask-entity correspondence poses a challenge to obtaining reliable pairs. To address this, we employ a vision-language large model to re-caption the input images and extract precise entities, and we design a multi-scale matching strategy to reduce noisy mask-entity pairs. Our Unpair-Seg framework demonstrates impressive performance, achieving 14.6\% and 19.5\% mIoU on the ADE-847 and PASCAL Context-459 datasets, significantly narrowing the gap between fully-supervised and weakly-supervised methods.

Haodong Chen, Runnan Chen, Qiang Qu et al.

Recent advancements in 3D Gaussian Splatting (3DGS) have substantially improved novel view synthesis, enabling high-quality reconstruction and real-time rendering. However, blurring artifacts, such as floating primitives and over-reconstruction, remain challenging. Current methods address these issues by refining scene structure, enhancing geometric representations, addressing blur in training images, improving rendering consistency, and optimizing density control, yet the role of kernel design remains underexplored. We identify the soft boundaries of Gaussian ellipsoids as one of the causes of these artifacts, limiting detail capture in high-frequency regions. To bridge this gap, we introduce 3D Linear Splatting (3DLS), which replaces Gaussian kernels with linear kernels to achieve sharper and more precise results, particularly in high-frequency regions. Through evaluations on three datasets, 3DLS demonstrates state-of-the-art fidelity and accuracy, along with a 30% FPS improvement over baseline 3DGS. The implementation will be made publicly available upon acceptance.

Zhenchen Wan, Yanwu Xu, Zhaoqing Wang et al.

Recent advancements in Virtual Try-On (VTO) have demonstrated exceptional efficacy in generating realistic images and preserving garment details, largely attributed to the robust generative capabilities of text-to-image (T2I) diffusion backbones. However, the T2I models that underpin these methods have become outdated, thereby limiting the potential for further improvement in VTO. Additionally, current methods face notable challenges in accurately rendering text on garments without distortion and preserving fine-grained details, such as textures and material fidelity. The emergence of Diffusion Transformer (DiT) based T2I models has showcased impressive performance and offers a promising opportunity for advancing VTO. Directly applying existing VTO techniques to transformer-based T2I models is ineffective due to substantial architectural differences, which hinder their ability to fully leverage the models' advanced capabilities for improved text generation. To address these challenges and unlock the full potential of DiT-based T2I models for VTO, we propose TED-VITON, a novel framework that integrates a Garment Semantic (GS) Adapter for enhancing garment-specific features, a Text Preservation Loss to ensure accurate and distortion-free text rendering, and a constraint mechanism to generate prompts by optimizing Large Language Model (LLM). These innovations enable state-of-the-art (SOTA) performance in visual quality and text fidelity, establishing a new benchmark for VTO task. Project page: https://zhenchenwan.github.io/TED-VITON/

Jiepeng Wang, Zhaoqing Wang, Hao Pan et al.

A unified diffusion framework for multi-modal generation and understanding has the transformative potential to achieve seamless and controllable image diffusion and other cross-modal tasks. In this paper, we introduce MMGen, a unified framework that integrates multiple generative tasks into a single diffusion model. This includes: (1) multi-modal category-conditioned generation, where multi-modal outputs are generated simultaneously through a single inference process, given category information; (2) multi-modal visual understanding, which accurately predicts depth, surface normals, and segmentation maps from RGB images; and (3) multi-modal conditioned generation, which produces corresponding RGB images based on specific modality conditions and other aligned modalities. Our approach develops a novel diffusion transformer that flexibly supports multi-modal output, along with a simple modality-decoupling strategy to unify various tasks. Extensive experiments and applications demonstrate the effectiveness and superiority of MMGen across diverse tasks and conditions, highlighting its potential for applications that require simultaneous generation and understanding.

Zhenchen Wan, Yanwu xu, Dongting Hu et al.

Recent advancements in Virtual Try-On (VITON) have significantly improved image realism and garment detail preservation, driven by powerful text-to-image (T2I) diffusion models. However, existing methods often rely on user-provided masks, introducing complexity and performance degradation due to imperfect inputs, as shown in Fig.1(a). To address this, we propose a Mask-Free VITON (MF-VITON) framework that achieves realistic VITON using only a single person image and a target garment, eliminating the requirement for auxiliary masks. Our approach introduces a novel two-stage pipeline: (1) We leverage existing Mask-based VITON models to synthesize a high-quality dataset. This dataset contains diverse, realistic pairs of person images and corresponding garments, augmented with varied backgrounds to mimic real-world scenarios. (2) The pre-trained Mask-based model is fine-tuned on the generated dataset, enabling garment transfer without mask dependencies. This stage simplifies the input requirements while preserving garment texture and shape fidelity. Our framework achieves state-of-the-art (SOTA) performance regarding garment transfer accuracy and visual realism. Notably, the proposed Mask-Free model significantly outperforms existing Mask-based approaches, setting a new benchmark and demonstrating a substantial lead over previous approaches. For more details, visit our project page: https://zhenchenwan.github.io/MF-VITON/.

Xiaoli Wei, Zhaoqing Wang, Yandong Guo et al.

Interactive video object segmentation is a crucial video task, having various applications from video editing to data annotating. However, current approaches struggle to accurately segment objects across diverse domains. Recently, Segment Anything Model (SAM) introduces interactive visual prompts and demonstrates impressive performance across different domains. In this paper, we propose a training-free prompt tracking framework for interactive video object segmentation (I-PT), leveraging the powerful generalization of SAM. Although point tracking efficiently captures the pixel-wise information of objects in a video, points tend to be unstable when tracked over a long period, resulting in incorrect segmentation. Towards fast and robust interaction, we jointly adopt sparse points and boxes tracking, filtering out unstable points and capturing object-wise information. To better integrate reference information from multiple interactions, we introduce a cross-round space-time module (CRSTM), which adaptively aggregates mask features from previous rounds and frames, enhancing the segmentation stability. Our framework has demonstrated robust zero-shot video segmentation results on popular VOS datasets with interaction types, including DAVIS 2017, YouTube-VOS 2018, and MOSE 2023, maintaining a good tradeoff between performance and interaction time.

Zhaoqing Wang, Yu Lu, Qiang Li et al.

Referring image segmentation aims to segment a referent via a natural linguistic expression.Due to the distinct data properties between text and image, it is challenging for a network to well align text and pixel-level features. Existing approaches use pretrained models to facilitate learning, yet separately transfer the language/vision knowledge from pretrained models, ignoring the multi-modal corresponding information. Inspired by the recent advance in Contrastive Language-Image Pretraining (CLIP), in this paper, we propose an end-to-end CLIP-Driven Referring Image Segmentation framework (CRIS). To transfer the multi-modal knowledge effectively, CRIS resorts to vision-language decoding and contrastive learning for achieving the text-to-pixel alignment. More specifically, we design a vision-language decoder to propagate fine-grained semantic information from textual representations to each pixel-level activation, which promotes consistency between the two modalities. In addition, we present text-to-pixel contrastive learning to explicitly enforce the text feature similar to the related pixel-level features and dissimilar to the irrelevances. The experimental results on three benchmark datasets demonstrate that our proposed framework significantly outperforms the state-of-the-art performance without any post-processing. The code will be released.

Zhaoqing Wang, Qiang Li, Guoxin Zhang et al.

By considering the spatial correspondence, dense self-supervised representation learning has achieved superior performance on various dense prediction tasks. However, the pixel-level correspondence tends to be noisy because of many similar misleading pixels, e.g., backgrounds. To address this issue, in this paper, we propose to explore \textbf{set} \textbf{sim}ilarity (SetSim) for dense self-supervised representation learning. We generalize pixel-wise similarity learning to set-wise one to improve the robustness because sets contain more semantic and structure information. Specifically, by resorting to attentional features of views, we establish corresponding sets, thus filtering out noisy backgrounds that may cause incorrect correspondences. Meanwhile, these attentional features can keep the coherence of the same image across different views to alleviate semantic inconsistency. We further search the cross-view nearest neighbours of sets and employ the structured neighbourhood information to enhance the robustness. Empirical evaluations demonstrate that SetSim is superior to state-of-the-art methods on object detection, keypoint detection, instance segmentation, and semantic segmentation.