Yolov5-jetson

yolov5 TensorRT implementation for running on Nvidia Jetson AGX Xavier with RealSense D435.

This repo uses yolov5 release v3.0.

Acknowledgement

This repo is a modified version of https://github.com/wang-xinyu/tensorrtx/tree/master/yolov5.

The motivation is that the origin python implementation for yolov5 inference with TensorRT acceleration does not work on my Nvidia Jetson Xavier. Therefore, I use Pybind11 to add a Python interface for the C++ implementation.

Config

The configuration approach is not well-designed, I will consider refactoring when I have time. Currently I just bare with the version from the original repo.

• Choose the model s/m/l/x by NET macro in yolov5.cpp
• Input shape defined in yololayer.h
• Number of classes defined in yololayer.h
• FP16/FP32 can be selected by the macro in yolov5.cpp
• GPU id can be selected by the macro in yolov5.cpp
• NMS thresh in yolov5.cpp
• BBox confidence thresh in yolov5.cpp
• Batch size in yolov5.cpp

Usage (Yolov5s as an example)

1. Generate .wts from pytorch with .pt, or download .wts from model zoo

   • git clone source code of yolov5 v3.0
   • download https://github.com/ultralytics/yolov5/releases/download/v3.0/yolov5s.pt
   • copy scripts/gen_wts.py into ultralytics/yolov5
   • ensure the file name is yolov5s.pt and yolov5s.wts in gen_wts.py
   • go to this repo folder
   • Execute python3 gen_wts.py
   • a file 'yolov5s.wts' will be generated

2. Build this repo and run

   • Put yolov5s.wts into this repo folder

   • Update CLASS_NUM in yololayer.h if your model is trained on custom dataset

   • Execute the following bash commands

     mkdir build
     cd build
     cmake ..
     make
     // serialize model to plan file
     sudo ./yolov5 -s [.wts] [.engine] [s/m/l/x or c gd gw]
     // deserialize and run inference, the images in [image folder] will be processed.
     sudo ./yolov5 -d [.engine] [image folder]
     // For example yolov5s
     sudo ./yolov5 -s yolov5s.wts yolov5s.engine s
     sudo ./yolov5 -d yolov5s.engine ../samples
     // For example Custom model with depth_multiple=0.17, width_multiple=0.25 in yolov5.yaml
     sudo ./yolov5 -s yolov5_custom.wts yolov5.engine c 0.17 0.25
     sudo ./yolov5 -d yolov5.engine ../samples

3. check the images generated.

4. For inference with python, an example is given in scripts/yolov5_infer.py