This repository serves as a comprehensive collection of research papers and projects focused on Bird's Eye View (BEV) perception for autonomous driving. It aims to provide a centralized resource for researchers and engineers working on tasks such as 3D object detection, segmentation, mapping, and occupancy prediction, offering a broad overview of the state-of-the-art in this domain.

How It Works

The repository categorizes a vast number of academic papers based on their primary task (e.g., 3D Object Detection, BEV Segmentation, Mapping, Occupancy Prediction) and the sensor modalities used (e.g., LiDAR, Camera, Radar, Multi-modal). It acts as a curated index, linking to papers and their associated GitHub repositories or project pages, facilitating discovery and access to relevant research.

Highlighted Details

• Extensive coverage of 3D object detection methods using various sensor combinations (Radar-Lidar, Radar-Camera, Lidar-Camera, Monocular, Multi-camera).
• Detailed sections on BEV segmentation, mapping, and occupancy prediction, showcasing diverse approaches and datasets.
• Includes papers on perception, prediction, and planning, reflecting the end-to-end pipeline in autonomous driving.
• Features a "World Model" section, highlighting research on generative models and large language models for driving scenarios.

Maintenance & Community

The repository was initially updated in May 2023, with subsequent additions of papers. It appears to be a static collection of research links rather than an actively maintained project with a dedicated community or roadmap.

Licensing & Compatibility

As this repository is a collection of links to external research papers and projects, it does not have its own licensing. The licensing of individual linked projects would need to be checked separately.

Limitations & Caveats

The repository is a curated list of papers and does not provide code, datasets, or pre-trained models directly. Users must follow the links to access the actual research implementations and resources. The sheer volume of papers may require significant effort to navigate and evaluate.