Core Viewpoint - Hesai (HSAI.US) has achieved a significant milestone by becoming the first company globally to produce over one million LiDAR units annually, marking a pivotal moment in its vision to enable 1% of vehicles worldwide with 3D perception capabilities by 2025 [1] Company Summary - Hesai's stock rose over 7% pre-market, reaching $23.62 following the announcement of its one millionth LiDAR unit produced by the end of September [1] - The company aims to lead the LiDAR industry with its "million units per year" production capacity, setting new standards in the sector [1] Industry Summary - Morgan Stanley has raised its forecast for 2025 automobile sales in mainland China by 6% to 29.9 million units, indicating a 9% year-on-year increase [1] - The anticipated surge in vehicle purchases is attributed to the expected pre-purchase wave before the expiration of stimulus policies and the launch of several new models, which are expected to boost wholesale sales in the fourth quarter [1] - Morgan Stanley has also increased Hesai's target price to $35, maintaining an "Overweight" rating [1]

Core Viewpoint - The article discusses the rapid industrialization of Feed-Forward 3D Gaussian Splatting (3DGS) and introduces VolSplat, which abandons the traditional pixel-aligned strategy in favor of a voxel-aligned framework, enhancing robustness, efficiency, and engineering feasibility in multi-view rendering [1][2]. Summary by Sections Introduction to VolSplat - VolSplat addresses the limitations of existing pixel-aligned methods, which struggle with precise alignment of 2D features in 3D space and are constrained by the pixel grid in Gaussian density allocation [2][6]. Performance Comparison - Experimental results on public datasets like RealEstate10K and ScanNet show that VolSplat outperforms various pixel-aligned baselines in visual quality and geometric consistency [4][5]. Core Concepts of VolSplat - The core idea of VolSplat is to shift alignment from 2D to 3D, allowing for better integration of multi-view information and overcoming challenges related to multi-view consistency and Gaussian density allocation [6][9]. Methodology Breakdown - The VolSplat pipeline consists of three clear modules: 1. 2D feature extraction and depth estimation 2. Lifting pixels to voxels and feature aggregation 3. Sparse 3D refinement and Gaussian regression [9][11]. Step-by-Step Process - **Step 1**: 2D features are extracted using a shared encoder, and depth maps are constructed to provide necessary geometric priors for subsequent processing [11]. - **Step 2**: Pixels are projected into 3D space based on predicted depths, creating a point cloud that is voxelized for feature aggregation, enhancing cross-view consistency [12][13]. - **Step 3**: A sparse 3D U-Net refines voxel features, predicting corrections for each voxel and regressing Gaussian parameters for rendering [14]. Experimental Highlights - VolSplat demonstrates superior zero-shot generalization across datasets, maintaining high performance even on unseen data, with a PSNR of 32.65 dB on the ACID dataset [15][17]. Practical Implications - The advancements in VolSplat lead to fewer artifacts and better geometric fidelity, translating to improved user experiences in applications like virtual tours and indoor navigation [17][19]. Future Directions - VolSplat opens new avenues for research in 3D reconstruction, robotics, autonomous driving, and AR/VR, providing a unified framework for integrating multimodal data [19][20].

Core Insights - A new "dual-modal" lidar system has been developed by Huazhong University of Science and Technology in collaboration with Tsinghua University and Beijing Information Science and Technology University, enhancing the 3D perception capabilities for autonomous vehicles, robots, and drones [1] Group 1 - The innovative lidar system aims to improve the three-dimensional sensing abilities of various devices [1] - The research findings have been published in the international optical journal "Light: Science & Applications" [1]

Core Viewpoint - Chinese scientists have successfully developed a new "dual-mode" LiDAR system that enhances the three-dimensional perception capabilities of autonomous vehicles, robots, and drones [1][4]. Group 1: Technology Development - The "dual-mode" LiDAR system was developed by a team from Huazhong University of Science and Technology in collaboration with Tsinghua University and Beijing Information Science and Technology University, and published in the top international optical journal "Light: Science & Applications" [1]. - Traditional LiDAR systems are categorized into scanning and flash types, with the former offering high precision but slow speed, while the latter provides fast speed but limited precision. The new system aims to combine the advantages of both [1][4]. Group 2: System Functionality - The core of the new system is a device called "hybrid cascaded translational metasurface," which consists of two layers of specially designed nano-lenses. By altering the polarization state of the incident laser, the system can switch between high-precision scanning mode and flash mode [1][2]. - The hardware allows the system to automatically switch modes based on actual needs, first using flash mode to quickly gather environmental information, then switching to scanning mode for detailed detection of specific areas [2]. Group 3: Implications and Future Applications - This technology successfully integrates the advantages of traditional scanning and flash LiDAR, ensuring high precision and long-distance detection while improving detection efficiency [4]. - The more efficient, flexible, and compact "dual-mode" LiDAR is expected to play a significant role in the fields of autonomous vehicles, intelligent robots, and drones, enhancing their ability to perceive the world [4].