Zeyun Zhong

Zeyun Zhong, Manuel Martin, Chengzhi Wu et al.

Recent Large Multimodal Models (LMMs), primarily designed for offline settings, are ill-suited for the dynamic requirements of streaming video. While recent online adaptations improve real-time processing, they still face critical scalability challenges, with resource demands typically growing at least linearly with video duration. To overcome this bottleneck, we propose FlowNar, a novel framework for scalable streaming video narration. The core of FlowNar is a dynamic context management strategy for historical visual context removal, combined with our CLAM (Cross Linear Attentive Memory) module for streaming visual history retention, ensuring bounded visual memory usage and computational complexity, crucial for efficient streaming. We also introduce a realistic self-conditioned evaluation protocol and complementary evaluation metrics to assess streaming narration models under deployment-like conditions. Experiments on the Ego4D, EgoExo4D, and EpicKitchens100 datasets demonstrate that FlowNar substantially improves narration quality over strong baselines while being highly efficient, supporting processing of 10$\times$ longer videos and achieving 3$\times$ higher throughput (FPS). The code is available at https://github.com/zeyun-zhong/FlowNar.

Di Wen, Zeyun Zhong, David Schneider et al.

We introduce IMPACT, a synchronized five-view RGB-D dataset for deployment-oriented industrial procedural understanding, built around real assembly and disassembly of a commercial angle grinder with professional-grade tools. To our knowledge, IMPACT is the first real industrial assembly benchmark that jointly provides synchronized ego-exo RGB-D capture, decoupled bimanual annotation, compliance-aware state tracking, and explicit anomaly--recovery supervision within a single real industrial workflow. It comprises 112 trials from 13 participants totaling 39.5 hours, with multi-route execution governed by a partial-order prerequisite graph, a six-category anomaly taxonomy, and operator cognitive load measured via NASA-TLX. The annotation hierarchy links hand-specific atomic actions to coarse procedural steps, component assembly states, and per-hand compliance phases, with synchronized null spans across views to decouple perceptual limitations from algorithmic failure. Systematic baselines reveal fundamental limitations that remain invisible to single-task benchmarks, particularly under realistic deployment conditions that involve incomplete observations, flexible execution paths, and corrective behavior. The full dataset, annotations, and evaluation code are available at https://github.com/Kratos-Wen/IMPACT.

Zeyun Zhong, David Schneider, Michael Voit et al.

Although human action anticipation is a task which is inherently multi-modal, state-of-the-art methods on well known action anticipation datasets leverage this data by applying ensemble methods and averaging scores of unimodal anticipation networks. In this work we introduce transformer based modality fusion techniques, which unify multi-modal data at an early stage. Our Anticipative Feature Fusion Transformer (AFFT) proves to be superior to popular score fusion approaches and presents state-of-the-art results outperforming previous methods on EpicKitchens-100 and EGTEA Gaze+. Our model is easily extensible and allows for adding new modalities without architectural changes. Consequently, we extracted audio features on EpicKitchens-100 which we add to the set of commonly used features in the community.

Weitong Kong, Di Wen, Kunyu Peng et al.

Correcting errors in long-video understanding is disproportionately costly: existing multimodal pipelines produce opaque, end-to-end outputs that expose no intermediate state for inspection, forcing annotators to revisit raw video and reconstruct temporal logic from scratch. The core bottleneck is not generation quality alone, but the absence of a supervisory interface through which human effort can be proportional to the scope of each error. We present IMPACT-CYCLE, a supervisory multi-agent system that reformulates long-video understanding as iterative claim-level maintenance of a shared semantic memory -- a structured, versioned state encoding typed claims, a claim dependency graph, and a provenance log. Role-specialized agents operating under explicit authority contracts decompose verification into local object-relation correctness, cross-temporal consistency, and global semantic coherence, with corrections confined to structurally dependent claims. When automated evidence is insufficient, the system escalates to human arbitration as the supervisory authority with final override rights; dependency-closure re-verification then ensures correction cost remains proportional to error scope. Experiments on VidOR show substantially improved downstream reasoning (VQA: 0.71 to 0.79) and a 4.8x reduction in human arbitration cost, with workload significantly lower than manual annotation. Code will be released at https://github.com/MKong17/IMPACT_CYCLE.

David J. Lerch, Livien Majer, Zeyun Zhong et al.

Robust self-supervised learning of multi-modal video representations is critical for real-world applications such as driver distraction detection, where multiple sensors provide complementary but noisy signals. Conventional contrastive objectives, such as InfoNCE, assume all negatives are equally informative and all positives are reliable. However, this assumption is frequently violated in multi-modal data due to viewpoint changes, occlusions, or semantic overlap across modalities. In this work, we propose a novel framework for multi-modal global alignment that addresses these challenges by jointly modeling faulty negatives and unreliable or faulty positives. We introduce soft targets derived from cycle-consistency scores to relax the hard-negative assumption, and a weighting mechanism based on similarity distributions to mitigate the impact of noisy or faulty positives. Our approach extends traditional pairwise alignment to a principled global multi-modal setting, aggregating alignment information across all modality pairs. We evaluate our method on the Drive&Act dataset, demonstrating that it consistently outperforms both pairwise and existing global alignment baselines across RGB, IR, Depth, and Skeleton modalities. Cross-view ablation studies further show strong generalization to unseen camera perspectives, highlighting the robustness of our representations. Overall, our framework provides a scalable and effective solution for self-supervised global multi-modal representation learning, enabling reliable driver distraction detection and pioneering in real-world multi-modal video understanding. Our code will be published on GitHub.

Zeyun Zhong, Manuel Martin, Michael Voit et al.

The ability to anticipate possible future human actions is essential for a wide range of applications, including autonomous driving and human-robot interaction. Consequently, numerous methods have been introduced for action anticipation in recent years, with deep learning-based approaches being particularly popular. In this work, we review the recent advances of action anticipation algorithms with a particular focus on daily-living scenarios. Additionally, we classify these methods according to their primary contributions and summarize them in tabular form, allowing readers to grasp the details at a glance. Furthermore, we delve into the common evaluation metrics and datasets used for action anticipation and provide future directions with systematical discussions.

Zeyun Zhong, Chengzhi Wu, Manuel Martin et al.

Anticipating future actions is inherently uncertain. Given an observed video segment containing ongoing actions, multiple subsequent actions can plausibly follow. This uncertainty becomes even larger when predicting far into the future. However, the majority of existing action anticipation models adhere to a deterministic approach, neglecting to account for future uncertainties. In this work, we rethink action anticipation from a generative view, employing diffusion models to capture different possible future actions. In this framework, future actions are iteratively generated from standard Gaussian noise in the latent space, conditioned on the observed video, and subsequently transitioned into the action space. Extensive experiments on four benchmark datasets, i.e., Breakfast, 50Salads, EpicKitchens, and EGTEA Gaze+, are performed and the proposed method achieves superior or comparable results to state-of-the-art methods, showing the effectiveness of a generative approach for action anticipation. Our code and trained models will be published on GitHub.

Haoshen Zhang, Di Wen, Kunyu Peng et al.

We present IMPACT-HOI, a mixed-initiative framework for annotating egocentric procedural video by constructing structured event graphs for Human-Object Interactions (HOI), motivated by the need for high-quality structured supervision for learning robot manipulation from human demonstration. IMPACT-HOI frames this task as the incremental resolution of a partially specified, onset-anchored event state. A trust-calibrated controller selects among direct queries, human-confirmed suggestions, and conservative completions based on empirical annotator behavior and evidence quality. A risk-bounded execution protocol, utilizing atomic rollback, ensures that human-confirmed decisions are preserved against conflicting automated updates. A user study with 9 participants shows a 13.5% reduction in manual annotation actions, a 46.67% event match rate, and zero confirmed-field violations under the studied protocol. The code will be made publicly available at https://github.com/541741106/IMPACT_HOI.

Qian Yin, Di Wen, Kunyu Peng et al.

Dense temporal annotation of procedural activity videos is vital for action understanding and embodied intelligence but remains labor-intensive due to reactive tools. Each correction is treated as an isolated edit, limiting reuse of information on annotator uncertainty and model reliability. We introduce IMPACT-Scribe, a correction-driven framework for dense labeling that uses each correction to improve future human-machine collaboration. IMPACT-Scribe combines uncertainty-aware boundary scribble supervision, local proposal modeling, cost-aware query planning, structured propagation, and correction-driven adaptation. Experiments and a human study show that this closed-loop design improves labeling quality per effort, enhances boundary accuracy, and fosters better human-machine interaction over time. The code will be made publicly available at https://github.com/BanzQians/IMPACT_AS.

Chengzhi Wu, Kaige Wang, Zeyun Zhong et al.

In recent years, there have been significant advancements in applying attention mechanisms to point cloud analysis. However, attention module variants featured in various research papers often operate under diverse settings and tasks, incorporating potential training strategies. This heterogeneity poses challenges in establishing a fair comparison among these attention module variants. In this paper, we address this issue by rethinking and exploring attention module design within a consistent base framework and settings. Both global-based and local-based attention methods are studied, with a focus on the selection basis and scales of neighbors for local-based attention. Different combinations of aggregated local features and computation methods for attention scores are evaluated, ranging from the initial addition/concatenation-based approach to the widely adopted dot product-based method and the recently proposed vector attention technique. Various position encoding methods are also investigated. Our extensive experimental analysis reveals that there is no universally optimal design across diverse point cloud tasks. Instead, drawing from best practices, we propose tailored attention modules for specific tasks, leading to superior performance on point cloud classification and segmentation benchmarks.

Zeyun Zhong, Manuel Martin, Frederik Diederichs et al.

This report presents a novel Mamba-based encoder-decoder architecture, QueryMamba, featuring an integrated verb-noun interaction module that utilizes a statistical verb-noun co-occurrence matrix to enhance video action forecasting. This architecture not only predicts verbs and nouns likely to occur based on historical data but also considers their joint occurrence to improve forecast accuracy. The efficacy of this approach is substantiated by experimental results, with the method achieving second place in the Ego4D LTA challenge and ranking first in noun prediction accuracy.

Chengzhi Wu, Yuxin Wan, Hao Fu et al.

Driven by the increasing demand for accurate and efficient representation of 3D data in various domains, point cloud sampling has emerged as a pivotal research topic in 3D computer vision. Recently, learning-to-sample methods have garnered growing interest from the community, particularly for their ability to be jointly trained with downstream tasks. However, previous learning-based sampling methods either lead to unrecognizable sampling patterns by generating a new point cloud or biased sampled results by focusing excessively on sharp edge details. Moreover, they all overlook the natural variations in point distribution across different shapes, applying a similar sampling strategy to all point clouds. In this paper, we propose a Sparse Attention Map and Bin-based Learning method (termed SAMBLE) to learn shape-specific sampling strategies for point cloud shapes. SAMBLE effectively achieves an improved balance between sampling edge points for local details and preserving uniformity in the global shape, resulting in superior performance across multiple common point cloud downstream tasks, even in scenarios with few-point sampling.