Wenxuan Zhang

Deyao Zhu, Jun Chen, Kilichbek Haydarov et al.

Asking insightful questions is crucial for acquiring knowledge and expanding our understanding of the world. However, the importance of questioning has been largely overlooked in AI research, where models have been primarily developed to answer questions. With the recent advancements of large language models (LLMs) like ChatGPT, we discover their capability to ask high-quality questions when provided with a suitable prompt. This discovery presents a new opportunity to develop an automatic questioning system. In this paper, we introduce ChatCaptioner, a novel automatic-questioning method deployed in image captioning. Here, ChatGPT is prompted to ask a series of informative questions about images to BLIP-2, a strong vision question-answering model. By keeping acquiring new visual information from BLIP-2's answers, ChatCaptioner is able to generate more enriched image descriptions. We conduct human-subject evaluations on common image caption datasets such as COCO, Conceptual Caption, and WikiArt, and compare ChatCaptioner with BLIP-2 as well as ground truth. Our results demonstrate that ChatCaptioner's captions are significantly more informative, receiving three times as many votes from human evaluators for providing the most image information. Besides, ChatCaptioner identifies 53% more objects within the image than BLIP-2 alone measured by WordNet synset matching. Code is available at https://github.com/Vision-CAIR/ChatCaptioner

Wenxuan Zhang, Xiaodong Cun, Xuan Wang et al.

Generating talking head videos through a face image and a piece of speech audio still contains many challenges. ie, unnatural head movement, distorted expression, and identity modification. We argue that these issues are mainly because of learning from the coupled 2D motion fields. On the other hand, explicitly using 3D information also suffers problems of stiff expression and incoherent video. We present SadTalker, which generates 3D motion coefficients (head pose, expression) of the 3DMM from audio and implicitly modulates a novel 3D-aware face render for talking head generation. To learn the realistic motion coefficients, we explicitly model the connections between audio and different types of motion coefficients individually. Precisely, we present ExpNet to learn the accurate facial expression from audio by distilling both coefficients and 3D-rendered faces. As for the head pose, we design PoseVAE via a conditional VAE to synthesize head motion in different styles. Finally, the generated 3D motion coefficients are mapped to the unsupervised 3D keypoints space of the proposed face render, and synthesize the final video. We conducted extensive experiments to demonstrate the superiority of our method in terms of motion and video quality.

Yifei Yuan, Wenxuan Zhang, Yang Deng et al.

Social media plays a significant role in boosting the fashion industry, where a massive amount of fashion-related posts are generated every day. In order to obtain the rich fashion information from the posts, we study the task of social media fashion knowledge extraction. Fashion knowledge, which typically consists of the occasion, person attributes, and fashion item information, can be effectively represented as a set of tuples. Most previous studies on fashion knowledge extraction are based on the fashion product images without considering the rich text information in social media posts. Existing work on fashion knowledge extraction in social media is classification-based and requires to manually determine a set of fashion knowledge categories in advance. In our work, we propose to cast the task as a captioning problem to capture the interplay of the multimodal post information. Specifically, we transform the fashion knowledge tuples into a natural language caption with a sentence transformation method. Our framework then aims to generate the sentence-based fashion knowledge directly from the social media post. Inspired by the big success of pre-trained models, we build our model based on a multimodal pre-trained generative model and design several auxiliary tasks for enhancing the knowledge extraction. Since there is no existing dataset which can be directly borrowed to our task, we introduce a dataset consisting of social media posts with manual fashion knowledge annotation. Extensive experiments are conducted to demonstrate the effectiveness of our model.

Meng Lu, Ran Xu, Yi Fang et al.

While recent vision-language models (VLMs) demonstrate strong image understanding, their ability to "think with images", i.e., to reason through multi-step visual interactions, remains limited. We introduce VISTA-Gym, a scalable training environment for incentivizing tool-integrated visual reasoning capabilities in VLMs. VISTA-Gym unifies diverse real-world multimodal reasoning tasks (7 tasks from 13 datasets in total) with a standardized interface for visual tools (e.g., grounding, parsing), executable interaction loops, verifiable feedback signals, and efficient trajectory logging, enabling visual agentic reinforcement learning at scale. While recent VLMs exhibit strong text-only reasoning, both proprietary and open-source models still struggle with tool selection, invocation, and coordination. With VISTA-Gym, we train VISTA-R1 to interleave tool-use with agentic reasoning via multi-turn trajectory sampling and end-to-end reinforcement learning. Extensive experiments across 11 public reasoning-intensive VQA benchmarks show that VISTA-R1-8B outperforms state-of-the-art baselines with similar sizes by 9.51%-18.72%, demonstrating VISTA-Gym as an effective training ground to unlock the tool-integrated reasoning capabilities for VLMs.

Yanjie Li, Wenxuan Zhang, Xinqi Lyu et al.

Recently, text-to-image diffusion models have been widely used for style mimicry and personalized customization through methods such as DreamBooth and Textual Inversion. This has raised concerns about intellectual property protection and the generation of deceptive content. Recent studies, such as Glaze and Anti-DreamBooth, have proposed using adversarial noise to protect images from these attacks. However, recent purification-based methods, such as DiffPure and Noise Upscaling, have successfully attacked these latest defenses, showing the vulnerabilities of these methods. Moreover, present methods show limited transferability across models, making them less effective against unknown text-to-image models. To address these issues, we propose a novel anti-mimicry method, StyleGuard. We propose a novel style loss that optimizes the style-related features in the latent space to make it deviate from the original image, which improves model-agnostic transferability. Additionally, to enhance the perturbation's ability to bypass diffusion-based purification, we designed a novel upscale loss that involves ensemble purifiers and upscalers during training. Extensive experiments on the WikiArt and CelebA datasets demonstrate that StyleGuard outperforms existing methods in robustness against various transformations and purifications, effectively countering style mimicry in various models. Moreover, StyleGuard is effective on different style mimicry methods, including DreamBooth and Textual Inversion. The code is available at https://github.com/PolyLiYJ/StyleGuard.

Christian Schuessler, Wenxuan Zhang, Johanna Bräunig et al.

In the fast-paced field of human-computer interaction (HCI) and virtual reality (VR), automatic gesture recognition has become increasingly essential. This is particularly true for the recognition of hand signs, providing an intuitive way to effortlessly navigate and control VR and HCI applications. Considering increased privacy requirements, radar sensors emerge as a compelling alternative to cameras. They operate effectively in low-light conditions without capturing identifiable human details, thanks to their lower resolution and distinct wavelength compared to visible light. While previous works predominantly deploy radar sensors for dynamic hand gesture recognition based on Doppler information, our approach prioritizes classification using an imaging radar that operates on spatial information, e.g. image-like data. However, generating large training datasets required for neural networks (NN) is a time-consuming and challenging process, often falling short of covering all potential scenarios. Acknowledging these challenges, this study explores the efficacy of synthetic data generated by an advanced radar ray-tracing simulator. This simulator employs an intuitive material model that can be adjusted to introduce data diversity. Despite exclusively training the NN on synthetic data, it demonstrates promising performance when put to the test with real measurement data. This emphasizes the practicality of our methodology in overcoming data scarcity challenges and advancing the field of automatic gesture recognition in VR and HCI applications.

Wenxuan Zhang, Peng Hu

Effective Edge AI for space object detection (SOD) tasks that can facilitate real-time collision assessment and avoidance is essential with the increasing space assets in near-Earth orbits. In SOD, low Earth orbit (LEO) satellites must detect other objects with high precision and minimal delay. We explore an Edge AI solution based on deep-learning-based vision sensing for SOD tasks and propose a deep learning model based on Squeeze-and-Excitation (SE) layers, Vision Transformers (ViT), and YOLOv9 framework. We evaluate the performance of these models across various realistic SOD scenarios, demonstrating their ability to detect multiple satellites with high accuracy and very low latency.

Peng Hu, Wenxuan Zhang

The growing density of satellites in low-Earth orbit (LEO) presents serious challenges to space sustainability, primarily due to the increased risk of in-orbit collisions. Traditional ground-based tracking systems are constrained by latency and coverage limitations, underscoring the need for onboard, vision-based space object detection (SOD) capabilities. In this paper, we propose a novel satellite clustering framework that enables the collaborative execution of deep learning (DL)-based SOD tasks across multiple satellites. To support this approach, we construct a high-fidelity dataset simulating imaging scenarios for clustered satellite formations. A distance-aware viewpoint selection strategy is introduced to optimize detection performance, and recent DL models are used for evaluation. Experimental results show that the clustering-based method achieves competitive detection accuracy compared to single-satellite and existing approaches, while maintaining a low size, weight, and power (SWaP) footprint. These findings underscore the potential of distributed, AI-enabled in-orbit systems to enhance space situational awareness and contribute to long-term space sustainability.

Norah Alballa, Wenxuan Zhang, Ziquan Liu et al.

Decentralized collaborative learning under data heterogeneity and privacy constraints has rapidly advanced. However, existing solutions like federated learning, ensembles, and transfer learning, often fail to adequately serve the unique needs of clients, especially when local data representation is limited. To address this issue, we propose a novel framework called Query-based Knowledge Transfer (QKT) that enables tailored knowledge acquisition to fulfill specific client needs without direct data exchange. QKT employs a data-free masking strategy to facilitate communication-efficient query-focused knowledge transfer while refining task-specific parameters to mitigate knowledge interference and forgetting. Our experiments, conducted on both standard and clinical benchmarks, show that QKT significantly outperforms existing collaborative learning methods by an average of 20.91\% points in single-class query settings and an average of 14.32\% points in multi-class query scenarios. Further analysis and ablation studies reveal that QKT effectively balances the learning of new and existing knowledge, showing strong potential for its application in decentralized learning.