search

NVIDIA-AI-IOT / Lidar_AI_Solution

A project demonstrating Lidar related AI solutions, including three GPU accelerated Lidar/camera DL networks (PointPillars, CenterPoint, BEVFusion) and the related libs (cuPCL, 3D SparseConvolution, YUV2RGB, cuOSD,).
Other
1.33k stars 236 forks source link

why not use TensorRT for LidarBackbone model in CUDA-BEVFusion? #285

Open fsi-kawatawa opened 1 month ago

fsi-kawatawa commented 1 month ago

Hello, In CUDA-BEVFusion, LiDAR Backbone uses ONNX format files (lidar.backbone.xyz.onnx). Is there any reason why the TensorRT transformation is not performed as in other models (e.g. camera.backbone.onnx)?

hopef commented 1 month ago

This is just because the DDS (data dependence shape) feature is not supported in TensorRT(<=8.x).