Yun Fu

Yin Yu, Ghasedi Kamran, Wu HsiangTao et al.

Recently, a surge of face editing techniques have been proposed to employ the pretrained StyleGAN for semantic manipulation. To successfully edit a real image, one must first convert the input image into StyleGAN's latent variables. However, it is still challenging to find latent variables, which have the capacity for preserving the appearance of the input subject (e.g., identity, lighting, hairstyles) as well as enabling meaningful manipulations. In this paper, we present a method to expand the latent space of StyleGAN with additional content features to break down the trade-off between low-distortion and high-editability. Specifically, we proposed a two-branch model, where the style branch first tackles the entanglement issue by the sparse manipulation of latent codes, and the content branch then mitigates the distortion issue by leveraging the content and appearance details from the input image. We confirm the effectiveness of our method using extensive qualitative and quantitative experiments on real face editing and reconstruction tasks.

Yitian Zhang, Xu Ma, Yue Bai et al.

Vision foundation models are renowned for the generalization ability due to massive training data. Nevertheless, they demand tremendous training resources, and the training data is often inaccessible, e.g., CLIP, DINOv2, posing great challenges to developing derivatives that could facilitate the research. In this work, we offer a very simple and general solution, named \textit{Proteus}, to distill foundation models into smaller equivalents on ImageNet-1K without access to the original training data. Specifically, we remove the designs from conventional knowledge distillation settings that result in dataset bias and present three levels of training objectives, i.e., token, patch, and feature, to maximize the efficacy of knowledge transfer. In this manner, Proteus is trained at ImageNet-level costs with surprising ability, facilitating the accessibility of training foundation models for the broader research community. When leveraging DINOv2-g/14 as the teacher, Proteus-L/14 matches the performance of the Oracle method DINOv2-L/14 (142M training data) across 19 benchmarks and outperforms other vision foundation models including CLIP-L/14 (400M), OpenCLIP-L/14 (400M/2B) and SynCLR-L/14 (600M) with a significantly smaller training set of 1.2M images.

Jianglin Lu, Yi Xu, Huan Wang et al.

Latent graph inference (LGI) aims to jointly learn the underlying graph structure and node representations from data features. However, existing LGI methods commonly suffer from the issue of supervision starvation, where massive edge weights are learned without semantic supervision and do not contribute to the training loss. Consequently, these supervision-starved weights, which may determine the predictions of testing samples, cannot be semantically optimal, resulting in poor generalization. In this paper, we observe that this issue is actually caused by the graph sparsification operation, which severely destroys the important connections established between pivotal nodes and labeled ones. To address this, we propose to restore the corrupted affinities and replenish the missed supervision for better LGI. The key challenge then lies in identifying the critical nodes and recovering the corrupted affinities. We begin by defining the pivotal nodes as $k$-hop starved nodes, which can be identified based on a given adjacency matrix. Considering the high computational burden, we further present a more efficient alternative inspired by CUR matrix decomposition. Subsequently, we eliminate the starved nodes by reconstructing the destroyed connections. Extensive experiments on representative benchmarks demonstrate that reducing the starved nodes consistently improves the performance of state-of-the-art LGI methods, especially under extremely limited supervision (6.12% improvement on Pubmed with a labeling rate of only 0.3%).

Haichao Zhang, Yi Xu, Yun Fu

Trajectory prediction is a critical task in computer vision and autonomous systems, playing a key role in autonomous driving, robotics, surveillance, and virtual reality. Existing methods often rely on complete and noise-free observational data, overlooking the challenges associated with out-of-sight objects and the inherent noise in sensor data caused by limited camera coverage, obstructions, and the absence of ground truth for denoised trajectories. These limitations pose safety risks and hinder reliable prediction in real-world scenarios. In this extended work, we present advancements in Out-of-Sight Trajectory (OST), a novel task that predicts the noise-free visual trajectories of out-of-sight objects using noisy sensor data. Building on our previous research, we broaden the scope of Out-of-Sight Trajectory Prediction (OOSTraj) to include pedestrians and vehicles, extending its applicability to autonomous driving, robotics, surveillance, and virtual reality. Our enhanced Vision-Positioning Denoising Module leverages camera calibration to establish a vision-positioning mapping, addressing the lack of visual references, while effectively denoising noisy sensor data in an unsupervised manner. Through extensive evaluations on the Vi-Fi and JRDB datasets, our approach achieves state-of-the-art performance in both trajectory denoising and prediction, significantly surpassing previous baselines. Additionally, we introduce comparisons with traditional denoising methods, such as Kalman filtering, and adapt recent trajectory prediction models to our task, providing a comprehensive benchmark. This work represents the first initiative to integrate vision-positioning projection for denoising noisy sensor trajectories of out-of-sight agents, paving the way for future advances. The code and preprocessed datasets are available at github.com/Hai-chao-Zhang/OST

Xu Ma, Peize Sun, Haoyu Ma et al.

Autoregressive (AR) models, long dominant in language generation, are increasingly applied to image synthesis but are often considered less competitive than Diffusion-based models. A primary limitation is the substantial number of image tokens required for AR models, which constrains both training and inference efficiency, as well as image resolution. To address this, we present Token-Shuffle, a novel yet simple method that reduces the number of image tokens in Transformer. Our key insight is the dimensional redundancy of visual vocabularies in Multimodal Large Language Models (MLLMs), where low-dimensional visual codes from visual encoder are directly mapped to high-dimensional language vocabularies. Leveraging this, we consider two key operations: token-shuffle, which merges spatially local tokens along channel dimension to decrease the input token number, and token-unshuffle, which untangles the inferred tokens after Transformer blocks to restore the spatial arrangement for output. Jointly training with textual prompts, our strategy requires no additional pretrained text-encoder and enables MLLMs to support extremely high-resolution image synthesis in a unified next-token prediction way while maintaining efficient training and inference. For the first time, we push the boundary of AR text-to-image generation to a resolution of 2048x2048 with gratifying generation performance. In GenAI-benchmark, our 2.7B model achieves 0.77 overall score on hard prompts, outperforming AR models LlamaGen by 0.18 and diffusion models LDM by 0.15. Exhaustive large-scale human evaluations also demonstrate our prominent image generation ability in terms of text-alignment, visual flaw, and visual appearance. We hope that Token-Shuffle can serve as a foundational design for efficient high-resolution image generation within MLLMs.

Yizhou Wang, Kuan-Chuan Peng, Yun Fu

3D anomaly detection and localization is of great significance for industrial inspection. Prior 3D anomaly detection and localization methods focus on the setting that the testing data share the same category as the training data which is normal. However, in real-world applications, the normal training data for the target 3D objects can be unavailable due to issues like data privacy or export control regulation. To tackle these challenges, we identify a new task -- zero-shot 3D anomaly detection and localization, where the training and testing classes do not overlap. To this end, we design 3DzAL, a novel patch-level contrastive learning framework based on pseudo anomalies generated using the inductive bias from task-irrelevant 3D xyz data to learn more representative feature representations. Furthermore, we train a normalcy classifier network to classify the normal patches and pseudo anomalies and utilize the classification result jointly with feature distance to design anomaly scores. Instead of directly using the patch point clouds, we introduce adversarial perturbations to the input patch xyz data before feeding into the 3D normalcy classifier for the classification-based anomaly score. We show that 3DzAL outperforms the state-of-the-art anomaly detection and localization performance.

Haichao Zhang, Yun Fu

Token-based video representation has emerged as a promising approach for enabling large language models (LLMs) to interpret video content. However, existing token reduction techniques, such as pruning and merging, often disrupt essential positional embeddings and rely on continuous visual tokens sampled from nearby pixels with similar spatial-temporal locations. By removing only a small fraction of tokens, these methods still produce relatively lengthy continuous sequences, which falls short of the extreme compression required to balance computational efficiency and token count in video LLMs. In this paper, we introduce the novel task of Extreme Short Token Reduction, which aims to represent entire videos using a minimal set of discrete tokens. We propose VQToken, a neural discrete token representation framework that (i) applies adaptive vector quantization to continuous ViT embeddings to learn a compact codebook and (ii) preserves spatial-temporal positions via a token hash function by assigning each grid-level token to its nearest codebook entry. On the Extreme Short Token Reduction task, our VQToken compresses sequences to just 0.07 percent of their original length while incurring only a 0.66 percent drop in accuracy on the NextQA-MC benchmark. It also achieves comparable performance on ActNet-QA, Long Video Bench, and VideoMME. We further introduce the Token Information Density (TokDense) metric and formalize fixed-length and adaptive-length subtasks, achieving state-of-the-art results in both settings. Our approach dramatically lowers theoretical complexity, increases information density, drastically reduces token counts, and enables efficient video LLMs in resource-constrained environments.

Haichao Zhang, Wenhao Chai, Shwai He et al.

High temporal resolution is essential for capturing fine-grained details in video understanding. However, current video large language models (VLLMs) and benchmarks mostly rely on low-frame-rate sampling, such as uniform sampling or keyframe selection, discarding dense temporal information. This compromise avoids the high cost of tokenizing every frame, which otherwise leads to redundant computation and linear token growth as video length increases. While this trade-off works for slowly changing content, it fails for tasks like lecture comprehension, where information appears in nearly every frame and requires precise temporal alignment. To address this gap, we introduce Dense Video Understanding (DVU), which enables high-FPS video comprehension by reducing both tokenization time and token overhead. Existing benchmarks are also limited, as their QA pairs focus on coarse content changes. We therefore propose DIVE (Dense Information Video Evaluation), the first benchmark designed for dense temporal reasoning. To make DVU practical, we present Gated Residual Tokenization (GRT), a two-stage framework: (1) Motion-Compensated Inter-Gated Tokenization uses pixel-level motion estimation to skip static regions during tokenization, achieving sub-linear growth in token count and compute. (2) Semantic-Scene Intra-Tokenization Merging fuses tokens across static regions within a scene, further reducing redundancy while preserving dynamic semantics. Experiments on DIVE show that GRT outperforms larger VLLM baselines and scales positively with FPS. These results highlight the importance of dense temporal information and demonstrate that GRT enables efficient, scalable high-FPS video understanding.

Zhawnen Chen, Tianchun Wang, Yizhou Wang et al.

Can large multimodal models have a human-like ability for emotional and social reasoning, and if so, how does it work? Recent research has discovered emergent theory-of-mind (ToM) reasoning capabilities in large language models (LLMs). LLMs can reason about people's mental states by solving various text-based ToM tasks that ask questions about the actors' ToM (e.g., human belief, desire, intention). However, human reasoning in the wild is often grounded in dynamic scenes across time. Thus, we consider videos a new medium for examining spatio-temporal ToM reasoning ability. Specifically, we ask explicit probing questions about videos with abundant social and emotional reasoning content. We develop a pipeline for multimodal LLM for ToM reasoning using video and text. We also enable explicit ToM reasoning by retrieving key frames for answering a ToM question, which reveals how multimodal LLMs reason about ToM.

Joseph P. Robinson, Yu Yin, Zaid Khan et al.

Recognizing Families In the Wild (RFIW): an annual large-scale, multi-track automatic kinship recognition evaluation that supports various visual kin-based problems on scales much higher than ever before. Organized in conjunction with the 15th IEEE International Conference on Automatic Face and Gesture Recognition (FG) as a Challenge, RFIW provides a platform for publishing original work and the gathering of experts for a discussion of the next steps. This paper summarizes the supported tasks (i.e., kinship verification, tri-subject verification, and search & retrieval of missing children) in the evaluation protocols, which include the practical motivation, technical background, data splits, metrics, and benchmark results. Furthermore, top submissions (i.e., leader-board stats) are listed and reviewed as a high-level analysis on the state of the problem. In the end, the purpose of this paper is to describe the 2020 RFIW challenge, end-to-end, along with forecasts in promising future directions.

Pengyu Gao, Siyu Xia, Joseph Robinson et al.

Visual kinship recognition aims to identify blood relatives from facial images. Its practical application-- like in law-enforcement, video surveillance, automatic family album management, and more-- has motivated many researchers to put forth effort on the topic as of recent. In this paper, we focus on a new view of visual kinship technology: kin-based face generation. Specifically, we propose a two-stage kin-face generation model to predict the appearance of a child given a pair of parents. The first stage includes a deep generative adversarial autoencoder conditioned on ages and genders to map between facial appearance and high-level features. The second stage is our proposed DNA-Net, which serves as a transformation between the deep and genetic features based on a random selection process to fuse genes of a parent pair to form the genes of a child. We demonstrate the effectiveness of the proposed method quantitatively and qualitatively: quantitatively, pre-trained models and human subjects perform kinship verification on the generated images of children; qualitatively, we show photo-realistic face images of children that closely resemble the given pair of parents. In the end, experiments validate that the proposed model synthesizes convincing kin-faces using both subjective and objective standards.