{"id":24108984,"url":"https://github.com/knzo25/bevfusion_ros2","last_synced_at":"2025-03-15T23:38:11.171Z","repository":{"id":260435097,"uuid":"881289075","full_name":"knzo25/bevfusion_ros2","owner":"knzo25","description":"BEVFusion implementation in ROS2","archived":false,"fork":false,"pushed_at":"2025-01-27T05:02:00.000Z","size":199,"stargazers_count":11,"open_issues_count":0,"forks_count":1,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-01-27T06:18:47.526Z","etag":null,"topics":["3d-object-detection","autoware","autoware-developers","camera-lidar","machine-learning","ros2"],"latest_commit_sha":null,"homepage":"","language":"C++","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/knzo25.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2024-10-31T09:20:39.000Z","updated_at":"2025-01-27T05:02:03.000Z","dependencies_parsed_at":"2024-10-31T10:24:16.409Z","dependency_job_id":"362ed526-58e6-4aab-9ac2-23973d160147","html_url":"https://github.com/knzo25/bevfusion_ros2","commit_stats":null,"previous_names":["knzo25/bevfusion_ros2"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/knzo25%2Fbevfusion_ros2","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/knzo25%2Fbevfusion_ros2/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/knzo25%2Fbevfusion_ros2/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/knzo25%2Fbevfusion_ros2/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/knzo25","download_url":"https://codeload.github.com/knzo25/bevfusion_ros2/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":243806031,"owners_count":20350773,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["3d-object-detection","autoware","autoware-developers","camera-lidar","machine-learning","ros2"],"created_at":"2025-01-11T00:01:34.077Z","updated_at":"2025-03-15T23:38:11.165Z","avatar_url":"https://github.com/knzo25.png","language":"C++","funding_links":[],"categories":["C++"],"sub_categories":[],"readme":"\n# Build the image\n\n```\ndocker build --network=host --progress=plain -t bevfusion_docker .\n```\n\n# Start the container\n\n```\ndocker run --net host --gpus all -v $(pwd)/autoware_lidar_bevfusion:/workspace/autoware/src/autoware_lidar_bevfusion -v $(pwd)/autoware_tensorrt_common:/workspace/autoware/src/autoware_tensorrt_common -it bevfusion_docker\n```\n\n# Compile the package\n\n```\nsource /opt/ros/humble/setup.bash \ncd autoware\ncolcon build --symlink-install --continue-on-error --packages-up-to autoware_lidar_bevfusion --event-handlers console_direct+ --cmake-args -DCMAKE_VERBOSE_MAKEFILE=ON -DCMAKE_BUILD_TYPE=Release\n```\n\n# Launch the node\n\n```\nsource install/setup.bash\nros2 launch autoware_lidar_bevfusion lidar_bevfusion.launch.xml model_path:=/workspace/autoware/src/autoware_lidar_bevfusion/config\n```\n\n# Benchmark\n\nWe provide some scripts to test and benchmark the inference of BEVFusion\n\n```bash\n./build/autoware_lidar_bevfusion/build_camera_lidar_engine \\\n    /workspace/autoware/src/autoware_lidar_bevfusion/config/bevfusion_traveller59.onnx \\\n    /workspace/autoware/build/autoware_lidar_bevfusion/libautoware_tensorrt_plugins.so \\\n    /workspace/autoware/bevfusion_traveller59.engine 0\n./build/autoware_lidar_bevfusion/run_camera_lidar_engine \\\n    build/autoware_lidar_bevfusion/libautoware_tensorrt_plugins.so \\\n    bevfusion_traveller59.engine 0\n\n./build/autoware_lidar_bevfusion/build_camera_lidar_engine \\\n    /workspace/autoware/src/autoware_lidar_bevfusion/config/bevfusion_cl_traveller59.onnx \\\n    /workspace/autoware/build/autoware_lidar_bevfusion/libautoware_tensorrt_plugins.so \\\n    /workspace/autoware/bevfusion_cl_traveller59.engine 1\n./build/autoware_lidar_bevfusion/run_camera_lidar_engine \\\n    build/autoware_lidar_bevfusion/libautoware_tensorrt_plugins.so \\\n    bevfusion_cl_traveller59.engine 1\n```\n\n| Modailty | GPU     | Backend  | Precision | #Voxels | Time [ms] |\n|----------|---------|----------|-----------|---------|-----------|\n| L        | RTX3060 | Pytorch  | fp32      | 68k     | 412.3     |\n| CL       | RTX3060 | Pytorch  | fp32      | 68k     | 4103.3    |\n| L        | RTX3060 | TensorRT | fp32      | 68k     | 56.1      |\n| CL       | RTX3060 | TensorRT | fp32      | 68k     | 132.5     |\n| L        | RTX3090 | TensorRT | fp32      | 68k     | 25.6      |\n| CL       | RTX3090 | TensorRT | fp32      | 68k     | 56.0      |\n\nMemory usage (including preprocessing buffers):\n - L: 1428MB\n - CL: 3072MB\n\n\n# Notes\n\n - BEVFusion lidar only and camera-lidar are compatible, although the current code only allows for camera-lidar\n - Can not be integrated into autoware yet due to TensorRT 10\n - The logic to handle missing cameras has not been implemented\n - The logic to trigger detection has not been fully implemented (depending on the sensor setup, the lidar may not trigger the inference)\n - The results should improve a few milliseconds since the onnx has redundant operations\n - If the latency needs to be further reduced, fp16 is needed. The current code does not handle that, but is not difficult","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fknzo25%2Fbevfusion_ros2","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fknzo25%2Fbevfusion_ros2","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fknzo25%2Fbevfusion_ros2/lists"}