mirrors/mmdeploy
1
0
Fork 0
mirror of https://github.com/open-mmlab/mmdeploy.git synced 2025-01-14 08:09:43 +08:00
Code Issues Projects Releases Wiki Activity

Improve mmdet3d doc (#1394)

Browse Source 
* docs(mmdet3d): add trt version desc

* docs(mmdet3d): update
This commit is contained in:
tpoisonooo 2022-11-18 18:34:30 +08:00 committed by GitHub
parent 0da6059342
commit 938ef537a7
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 12 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
docs/en/04-supported-codebases/mmdet3d.md
View File

@ -27,11 +27,14 @@ python tools/deploy.py \
| Model | Task | OnnxRuntime | TensorRT | ncnn | PPLNN | OpenVINO | Model config |
| :----------: | :------------: | :---------: | :------: | :--: | :---: | :------: | :------------------------------------------------------------------------------------: |
| PointPillars | VoxelDetection | Y | Y | N | N | Y | [config](https://github.com/open-mmlab/mmdetection3d/blob/master/configs/pointpillars) |
| PointPillars | VoxelDetection | Y | Y\* | N | N | Y | [config](https://github.com/open-mmlab/mmdetection3d/blob/master/configs/pointpillars) |
## Reminder
1. mmdet3d models on **cu102+TRT8.4** can be visualized normally. For cuda-11 or TRT8.2 users, these issues should be checked
Voxel detection onnx model excludes model.voxelize layer and model post process, and you can use python api to call these func.
- [TRT8.2 assertion `is_tensor`](https://github.com/NVIDIA/TensorRT/issues/1541)
- [TRT8.4 output NaN](https://github.com/NVIDIA/TensorRT/issues/2338)
2. Voxel detection onnx model excludes model.voxelize layer and model post process, and you can use python api to call these func.
Example:

9
docs/zh_cn/04-supported-codebases/mmdet3d.md
View File

@ -27,11 +27,14 @@ python tools/deploy.py \
| Model | Task | OnnxRuntime | TensorRT | ncnn | PPLNN | OpenVINO | Model config |
| :----------: | :------------: | :---------: | :------: | :--: | :---: | :------: | :------------------------------------------------------------------------------------: |
| PointPillars | VoxelDetection | Y | Y | N | N | Y | [config](https://github.com/open-mmlab/mmdetection3d/blob/master/configs/pointpillars) |
| PointPillars | VoxelDetection | Y | Y\* | N | N | Y | [config](https://github.com/open-mmlab/mmdetection3d/blob/master/configs/pointpillars) |
## 注意事项
1. TensorRT 实测 **cu102+TRT8.4** 可以正常可视化，使用 cuda-11 或 TRT8.2 版本需关注这些 issue
体素检测 onnx 模型不包含 model.voxelize 层和模型后处理，可用 python api 来调这些函数。
- [TRT8.2 assertion `is_tensor`](https://github.com/NVIDIA/TensorRT/issues/1541)
- [TRT8.4 output NaN](https://github.com/NVIDIA/TensorRT/issues/2338)
2. 体素检测 onnx 模型不包含 model.voxelize 层和模型后处理，可用 python api 来调这些函数。
示例：