Yi Liu

Yuanxin Liu, Kun Ouyang, Haoning Wu et al.

Recent studies have shown that long chain-of-thought (CoT) reasoning can significantly enhance the performance of large language models (LLMs) on complex tasks. However, this benefit is yet to be demonstrated in the domain of video understanding, since most existing benchmarks lack the reasoning depth required to demonstrate the advantages of extended CoT chains. While recent efforts have proposed benchmarks aimed at video reasoning, the tasks are often knowledge-driven and do not rely heavily on visual content. To bridge this gap, we introduce VideoReasonBench, a benchmark designed to evaluate vision-centric, complex video reasoning. To ensure visual richness and high reasoning complexity, each video in VideoReasonBench depicts a sequence of fine-grained operations on a latent state that is only visible in part of the video. The questions evaluate three escalating levels of video reasoning skills: recalling observed visual information, inferring the content of latent states, and predicting information beyond the video. Under such task setting, models have to precisely recall multiple operations in the video, and perform step-by-step reasoning to get correct final answers for these questions. Using VideoReasonBench, we comprehensively evaluate 18 state-of-the-art multimodal LLMs (MLLMs), finding that most perform poorly on complex video reasoning, e.g., GPT-4o achieves only 6.9% accuracy, while the thinking-enhanced Gemini-2.5-Pro significantly outperforms others with 56.0% accuracy. Our investigations on "test-time scaling" further reveal that extended thinking budget, while offering none or minimal benefits on existing video benchmarks, is essential for improving the performance on VideoReasonBench.

Jiaming Guo, Xueyi Zou, Yuyi Chen et al.

In recent years, videos and images in 720p (HD), 1080p (FHD) and 4K (UHD) resolution have become more popular for display devices such as TVs, mobile phones and VR. However, these high resolution images cannot achieve the expected visual effect due to the limitation of the internet bandwidth, and bring a great challenge for super-resolution networks to achieve real-time performance. Following this challenge, we explore multiple efficient network designs, such as pixel-unshuffle, repeat upscaling, and local skip connection removal, and propose a fast and lightweight super-resolution network. Furthermore, by analyzing the applications of the idea of divide-and-conquer in super-resolution, we propose assembled convolutions which can adapt convolution kernels according to the input features. Experiments suggest that our method outperforms all the state-of-the-art efficient super-resolution models, and achieves optimal results in terms of runtime and quality. In addition, our method also wins the first place in NTIRE 2023 Real-Time Super-Resolution - Track 1 ($\times$2). The code will be available at https://gitee.com/mindspore/models/tree/master/research/cv/AsConvSR

Jeffrey Ichnowski, Yahav Avigal, Yi Liu et al.

High-speed motions in pick-and-place operations are critical to making robots cost-effective in many automation scenarios, from warehouses and manufacturing to hospitals and homes. However, motions can be too fast -- such as when the object being transported has an open-top, is fragile, or both. One way to avoid spills or damage, is to move the arm slowly. We propose an alternative: Grasp-Optimized Motion Planning for Fast Inertial Transport (GOMP-FIT), a time-optimizing motion planner based on our prior work, that includes constraints based on accelerations at the robot end-effector. With GOMP-FIT, a robot can perform high-speed motions that avoid obstacles and use inertial forces to its advantage. In experiments transporting open-top containers with varying tilt tolerances, whereas GOMP computes sub-second motions that spill up to 90% of the contents during transport, GOMP-FIT generates motions that spill 0% of contents while being slowed by as little as 0% when there are few obstacles, 30% when there are high obstacles and 45-degree tolerances, and 50% when there 15-degree tolerances and few obstacles. Videos and more at: https://berkeleyautomation.github.io/gomp-fit/.

Bo Hu, Bryan Seybold, Shan Yang et al.

We present a method to infer the 3D pose of mice, including the limbs and feet, from monocular videos. Many human clinical conditions and their corresponding animal models result in abnormal motion, and accurately measuring 3D motion at scale offers insights into health. The 3D poses improve classification of health-related attributes over 2D representations. The inferred poses are accurate enough to estimate stride length even when the feet are mostly occluded. This method could be applied as part of a continuous monitoring system to non-invasively measure animal health.

Bai Li, Youmin Zhang, Yakun Ouyang et al.

This paper focuses on automatic guided vehicle (AGV) trajectory planning in the presence of moving obstacles with known but complicated trajectories. In order to achieve good solution precision, optimality and unification, the concerned task should be formulated as an optimal control problem, and then discretized into a nonlinear programming (NLP) problem, which is numerically optimized thereafter. Without a near-feasible or near-optimal initial guess, the NLP-solving process is usually slow. With the purpose of accelerating the NLP solution, a search-based rough planning stage is added to generate appropriate initial guesses. Concretely, a continuous state space is formulated, which consists of Cartesian product of 2D configuration space and a time dimension. The rough trajectory is generated by a graph-search based planner, namely the A* algorithm. Herein, the nodes in the graph are constructed by discretizing the aforementioned continuous spatio-temporal space. Through this first-search-then-optimization framework, optimal solutions to unified trajectory planning problems can be obtained fast. Simulations have been conducted to verify the real-time performance of our proposal.

Liyang Liu, Teng Zhang, Yi Liu et al.

Bundle adjustment (BA) with parallax angle based feature parameterization has been shown to have superior performance over BA using inverse depth or XYZ feature forms. In this paper, we propose an improved version of the parallax BA algorithm (PMBA) by extending it to the manifold domain along with observation-ray based objective function. With this modification, the problem formulation faithfully mimics the projective nature in a camera's image formation, BA is able to achieve better convergence, accuracy and robustness. This is particularly useful in handling diverse outdoor environments and collinear motion modes. Capitalizing on these properties, we further propose a pose-graph simplification to PMBA, with significant dimensionality reduction. This pose-graph model is convex in nature, easy to solve and its solution can serve as a good initial guess to the original BA problem which is intrinsically non-convex. We provide theoretical proof that our global initialization strategy can guarantee a near-optimal solution. Using a series of experiments involving diverse environmental conditions and motions, we demonstrate PMBA's superior convergence performance in comparison to other BA methods. We also show that, without incremental initialization or via third-party information, our global initialization process helps to bootstrap the full BA successfully in various scenarios, sequential or out-of-order, including some datasets from the "Bundle Adjustment in the Large" database.

Yi Liu, Hongying Meng, Mohammad Rafiq Swash et al.

With the rapid development of augmented reality (AR) and virtual reality (VR) technology, human-computer interaction (HCI) has been greatly improved for gaming interaction of AR and VR control. The finger micro-gesture is one of the important interactive methods for HCI applications such as in the Google Soli and Microsoft Kinect projects. However, the progress in this research is slow due to the lack of high quality public available database. In this paper, holoscopic 3D camera is used to capture high quality micro-gesture images and a new unique holoscopic 3D micro-gesture (HoMG) database is produced. The principle of the holoscopic 3D camera is based on the fly viewing system to see the objects. HoMG database recorded the image sequence of 3 conventional gestures from 40 participants under different settings and conditions. For the purpose of micro-gesture recognition, HoMG has a video subset with 960 videos and a still image subset with 30635 images. Initial micro-gesture recognition on both subsets has been conducted using traditional 2D image and video features and popular classifiers and some encouraging performance has been achieved. The database will be available for the research communities and speed up the research in this area.