Chengzhi Wu

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.

Yebin Yang, Di Wen, Lei Qi et al.

Text-guided 3D motion editing has seen success in single-person scenarios, but its extension to multi-person settings is less explored due to limited paired data and the complexity of inter-person interactions. We introduce the task of multi-person 3D motion editing, where a target motion is generated from a source and a text instruction. To support this, we propose InterEdit3D, a new dataset with manual two-person motion change annotations, and a Text-guided Multi-human Motion Editing (TMME) benchmark. We present InterEdit, a synchronized classifier-free conditional diffusion model for TMME. It introduces Semantic-Aware Plan Token Alignment with learnable tokens to capture high-level interaction cues and an Interaction-Aware Frequency Token Alignment strategy using DCT and energy pooling to model periodic motion dynamics. Experiments show that InterEdit improves text-to-motion consistency and edit fidelity, achieving state-of-the-art TMME performance. The dataset and code will be released at https://github.com/YNG916/InterEdit.

Junwei Zheng, Ruiping Liu, Yufan Chen et al.

Panoramic images, capturing a 360° field of view (FoV), encompass omnidirectional spatial information crucial for scene understanding. However, it is not only costly to obtain training-sufficient dense-annotated panoramas but also application-restricted when training models in a close-vocabulary setting. To tackle this problem, in this work, we define a new task termed Open Panoramic Segmentation (OPS), where models are trained with FoV-restricted pinhole images in the source domain in an open-vocabulary setting while evaluated with FoV-open panoramic images in the target domain, enabling the zero-shot open panoramic semantic segmentation ability of models. Moreover, we propose a model named OOOPS with a Deformable Adapter Network (DAN), which significantly improves zero-shot panoramic semantic segmentation performance. To further enhance the distortion-aware modeling ability from the pinhole source domain, we propose a novel data augmentation method called Random Equirectangular Projection (RERP) which is specifically designed to address object deformations in advance. Surpassing other state-of-the-art open-vocabulary semantic segmentation approaches, a remarkable performance boost on three panoramic datasets, WildPASS, Stanford2D3D, and Matterport3D, proves the effectiveness of our proposed OOOPS model with RERP on the OPS task, especially +2.2% on outdoor WildPASS and +2.4% mIoU on indoor Stanford2D3D. The source code is publicly available at https://junweizheng93.github.io/publications/OPS/OPS.html.

Chengzhi Wu, Kanran Zhou, Jan-Philipp Kaiser et al.

To enable automatic disassembly of different product types with uncertain conditions and degrees of wear in remanufacturing, agile production systems that can adapt dynamically to changing requirements are needed. Machine learning algorithms can be employed due to their generalization capabilities of learning from various types and variants of products. However, in reality, datasets with a diversity of samples that can be used to train models are difficult to obtain in the initial period. This may cause bad performances when the system tries to adapt to new unseen input data in the future. In order to generate large datasets for different learning purposes, in our project, we present a Blender add-on named MotorFactory to generate customized mesh models of various motor instances. MotorFactory allows to create mesh models which, complemented with additional add-ons, can be further used to create synthetic RGB images, depth images, normal images, segmentation ground truth masks, and 3D point cloud datasets with point-wise semantic labels. The created synthetic datasets may be used for various tasks including motor type classification, object detection for decentralized material transfer tasks, part segmentation for disassembly and handling tasks, or even reinforcement learning-based robotics control or view-planning.

Chengzhi Wu, Junwei Zheng, Julius Pfrommer et al.

Point cloud sampling is a less explored research topic for this data representation. The most commonly used sampling methods are still classical random sampling and farthest point sampling. With the development of neural networks, various methods have been proposed to sample point clouds in a task-based learning manner. However, these methods are mostly generative-based, rather than selecting points directly using mathematical statistics. Inspired by the Canny edge detection algorithm for images and with the help of the attention mechanism, this paper proposes a non-generative Attention-based Point cloud Edge Sampling method (APES), which captures salient points in the point cloud outline. Both qualitative and quantitative experimental results show the superior performance of our sampling method on common benchmark tasks.

Chengzhi Wu, Xuelei Bi, Julius Pfrommer et al.

On robotics computer vision tasks, generating and annotating large amounts of data from real-world for the use of deep learning-based approaches is often difficult or even impossible. A common strategy for solving this problem is to apply simulation-to-reality (sim2real) approaches with the help of simulated scenes. While the majority of current robotics vision sim2real work focuses on image data, we present an industrial application case that uses sim2real transfer learning for point cloud data. We provide insights on how to generate and process synthetic point cloud data in order to achieve better performance when the learned model is transferred to real-world data. The issue of imbalanced learning is investigated using multiple strategies. A novel patch-based attention network is proposed additionally to tackle this problem.

Chengzhi Wu, Julius Pfrommer, Jürgen Beyerer et al.

We present an improved approach for 3D object detection in point cloud data based on the Frustum PointNet (F-PointNet). Compared to the original F-PointNet, our newly proposed method considers the point neighborhood when computing point features. The newly introduced local neighborhood embedding operation mimics the convolutional operations in 2D neural networks. Thus features of each point are not only computed with the features of its own or of the whole point cloud but also computed especially with respect to the features of its neighbors. Experiments show that our proposed method achieves better performance than the F-Pointnet baseline on 3D object detection tasks.

Chengzhi Wu, Julius Pfrommer, Mingyuan Zhou et al.

We propose a combined generative and contrastive neural architecture for learning latent representations of 3D volumetric shapes. The architecture uses two encoder branches for voxel grids and multi-view images from the same underlying shape. The main idea is to combine a contrastive loss between the resulting latent representations with an additional reconstruction loss. That helps to avoid collapsing the latent representations as a trivial solution for minimizing the contrastive loss. A novel dynamic switching approach is used to cross-train two encoders with a shared decoder. The switching approach also enables the stop gradient operation on a random branch. Further classification experiments show that the latent representations learned with our self-supervised method integrate more useful information from the additional input data implicitly, thus leading to better reconstruction and classification performance.

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.

Chengzhi Wu, Junwei Zheng, Julius Pfrommer et al.

Modeling a 3D volumetric shape as an assembly of decomposed shape parts is much more challenging, but semantically more valuable than direct reconstruction from a full shape representation. The neural network needs to implicitly learn part relations coherently, which is typically performed by dedicated network layers that can generate transformation matrices for each part. In this paper, we propose a VoxAttention network architecture for attention-based part assembly. We further propose a variant of using channel-wise part attention and show the advantages of this approach. Experimental results show that our method outperforms most state-of-the-art methods for the part relation-aware 3D shape modeling task.

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.

Ruiping Liu, Junwei Zheng, Yufan Chen et al.

Egocentric memory is widely used in embodied intelligence, but it may be insufficient for comprehensive spatial-temporal reasoning. Inspired by human recall from both field and observer perspectives, we introduce EgoExoMem, the first benchmark for cross-view memory reasoning over synchronized egocentric and exocentric videos. EgoExoMem contains $2.6K$ high-quality MCQs across eight temporal, spatial, and cross-view QA types. To support dual-view retrieval, we propose E$^2$-Select, a training-free frame selection method for synchronized ego-exo videos. It combines relevance-based budget allocation with per-view k-DPP sampling to handle view asymmetry and cross-view temporal consistency. Experiments show that ego and exo views provide complementary memory cues, while existing MLLMs remain far from solving the benchmark: the best model reaches only $55.3\%$. E$^2$-Select achieves state-of-the-art performance of $58.2\%$ over frame-selection and RAG-based memory baselines. Further analysis reveals systematic view-preference conflicts between question framing and answer grounding, underscoring the novelty and challenge of cross-view memory reasoning.

Chengzhi Wu, Linxi Qiu, Kanran Zhou et al.

In this paper, we develop a novel benchmark suite including both a 2D synthetic image dataset and a 3D synthetic point cloud dataset. Our work is a sub-task in the framework of a remanufacturing project, in which small electric motors are used as fundamental objects. Apart from the given detection, classification, and segmentation annotations, the key objects also have multiple learnable attributes with ground truth provided. This benchmark can be used for computer vision tasks including 2D/3D detection, classification, segmentation, and multi-attribute learning. It is worth mentioning that most attributes of the motors are quantified as continuously variable rather than binary, which makes our benchmark well-suited for the less explored regression tasks. In addition, appropriate evaluation metrics are adopted or developed for each task and promising baseline results are provided. We hope this benchmark can stimulate more research efforts on the sub-domain of object attribute learning and multi-task learning in the future.

Chengzhi Wu, Hao Fu, Jan-Philipp Kaiser et al.

The accurate estimation of 6D pose remains a challenging task within the computer vision domain, even when utilizing 3D point cloud data. Conversely, in the manufacturing domain, instances arise where leveraging prior knowledge can yield advancements in this endeavor. This study focuses on the disassembly of starter motors to augment the engineering of product life cycles. A pivotal objective in this context involves the identification and 6D pose estimation of bolts affixed to the motors, facilitating automated disassembly within the manufacturing workflow. Complicating matters, the presence of occlusions and the limitations of single-view data acquisition, notably when motors are placed in a clamping system, obscure certain portions and render some bolts imperceptible. Consequently, the development of a comprehensive pipeline capable of acquiring complete bolt information is imperative to avoid oversight in bolt detection. In this paper, employing the task of bolt detection within the scope of our project as a pertinent use case, we introduce a meticulously devised pipeline. This multi-stage pipeline effectively captures the 6D information with regard to all bolts on the motor, thereby showcasing the effective utilization of prior knowledge in handling this challenging task. The proposed methodology not only contributes to the field of 6D pose estimation but also underscores the viability of integrating domain-specific insights to tackle complex problems in manufacturing and automation.

Chengzhi Wu, Qianliang Huang, Kun Jin et al.

Contrastive learning is an essential method in self-supervised learning. It primarily employs a multi-branch strategy to compare latent representations obtained from different branches and train the encoder. In the case of multi-modal input, diverse modalities of the same object are fed into distinct branches. When using single-modal data, the same input undergoes various augmentations before being fed into different branches. However, all existing contrastive learning frameworks have so far only performed contrastive operations on the learned features at the final loss end, with no information exchange between different branches prior to this stage. In this paper, for point cloud unsupervised learning without the use of extra training data, we propose a Contrastive Cross-branch Attention-based framework for Point cloud data (termed PoCCA), to learn rich 3D point cloud representations. By introducing sub-branches, PoCCA allows information exchange between different branches before the loss end. Experimental results demonstrate that in the case of using no extra training data, the representations learned with our self-supervised model achieve state-of-the-art performances when used for downstream tasks on point clouds.

Ruiping Liu, Yufan Chen, Yuheng Zhang et al.

Situated reasoning often relies on active exploration, yet in many real-world scenarios such exploration is infeasible due to physical constraints of robots or safety concerns of visually impaired users. Given only a limited observation, can an agent mentally simulate a future trajectory toward a target situation and answer spatial what-if questions? We introduce WanderDream, the first large-scale dataset designed for the emulative simulation of mental exploration, enabling models to reason without active exploration. WanderDream-Gen comprises 15.8K panoramic videos across 1,088 real scenes from HM3D, ScanNet++, and real-world captures, depicting imagined trajectories from current viewpoints to target situations. WanderDream-QA contains 158K question-answer pairs, covering starting states, paths, and end states along each trajectory to comprehensively evaluate exploration-based reasoning. Extensive experiments with world models and MLLMs demonstrate (1) that mental exploration is essential for situated reasoning, (2) that world models achieve compelling performance on WanderDream-Gen, (3) that imagination substantially facilitates reasoning on WanderDream-QA, and (4) that WanderDream data exhibit remarkable transferability to real-world scenarios. The source code and all data will be released.

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.