物体检测模型yolov5s.pt v5.0（官方下载）转换的tnn模型推理报错

[GitHubLuGeng]

1. 环境（environment）

• Build OS and Version: Ubuntu20.04
• RunTime OS Version: Linux

• RunTime DEVICE: CPU

  2. Github版本

• branch：master

• commit(optional):

  3. 编译方式(compile method) CMake完整编译参数(full cmake arguments)

  4. 编译日志(build log)

  将日志粘贴在这里
  Paste log here or pastebin

  5. 详细描述bug 情况 (Describe the bug) 1.使用在线方式（https://convertmodel.com/）把onnx模型转换成tnn模型推理报错如下“运行日志”

2.使用docker本地转换会报错如下： root@computer:/opt/TNN/tools/convert2tnn# python3 converter.py onnx2tnn /workspace/yolov5s-sim.onnx -o . -v v1.0 -optimize -align ---------- convert model, please wait a moment ----------

Converter ONNX to TNN Model...

Converter ONNX to TNN check_onnx_dim...

Converter ONNX to TNN check_onnx_dim...

Converter ONNX to TNN model failed!

root@computer:/opt/TNN/tools/convert2tnn#

3.附件 yolov5s-sim.zip yolov5s-sim.opt.tnn.zip

yolov5s-sim.zip是onnx模型 yolov5s-sim.opt.tnn.zip是在线转换的tnn模型

6. 运行日志（runtime log） root@computer:~/project/TNN/examples/linux/x86/build_linux_native$ ./demo_x86_objectdetectoryolov5 -p ../../../../model/yolov5/yolov5s-sim.opt.tnnproto -m ../../../../model/yolov5/yolov5s-sim.opt.tnnmodel Please choose the source you want to detect: Face-detector is about to start, and the picrture is ../../../assets/004545.jpg E/tnn: InferOutputShape [File /home/lu/project/TNN/source/tnn/layer/permute_layer.cc][Line 47] Permute param got wrong size. E/tnn: InferOutputShape [File /home/lu/project/TNN/source/tnn/layer/permute_layer.cc][Line 47] Permute param got wrong size. E/tnn: InferOutputShape [File /home/lu/project/TNN/source/tnn/layer/permute_layer.cc][Line 47] Permute param got wrong size. E/tnn: InferOutputShape [File /home/lu/project/TNN/source/tnn/layer/permute_layer.cc][Line 47] Permute param got wrong size. E/tnn: Init [File /home/lu/project/TNN/source/tnn/layer/base_layer.cc][Line 57] InferOutputShape failed E/tnn: InitLayers [File /home/lu/project/TNN/source/tnn/core/default_network.cc][Line 318] Error Init layer Transpose_260 (err: 4096 or 0x1000) Predictor Initing failed, please check the option parameters Segmentation fault (core dumped)

7. 截图（Screenshots）

 将截图粘贴在这里
 Paste screenshorts here or pastebin

[1627180283]

convertmodel.com 的 tnn 转换工具暂时没有更新，还不支持直接转 yolo v5。 docker 和本地编译的转换工具是支持的，你 docker 转换的时候在转换命令里加上 -debug 看一下是什么问题。

[GitHubLuGeng]

找到原因了，没有编译onnx2tnn cd /tools/convert2tnn ./build.sh