tensorrt_yolov7

Yolov7 TensorRT cpp

Description

This is a yolov7 TensorRT cpp app. Fisrt, using trtexec to convert onnx model to FP32 or FP16 TensorRT engine ,or INT8 TensorRT engine from the QAT model finetuned from yolov7_qat. Then you can use the detect/video_detect app to detect a list of images(images number must smaller than the batchsize of the model)/video. or use validate_coco app to test mAP of the TensorRT engine.

Prerequisites

Install opencv

• Note: There are OpenCV4 dependencies in this program. Follow README and documents of this repository https://github.com/opencv/opencv to install OpenCV. And, if you want use detect_video app, please install opencv with ffmpeg enabled

Install jsoncpp libs

jsoncpp lib is used to write coco-dataset-validate-result to json file.

$ sudo apt-get install libjsoncpp-dev

Build and Run yolov7-TensorRT-app

Build

$ mkdir build && cd build
$ cmake ..
$ make -j4

Prepare TensorRT engines

convert onnx model to tensorrt-engine

# fp32 model
$ /usr/src/tensorrt/bin/trtexec --onnx=yolov7.onnx --saveEngine=yolov7fp32.engine
# fp16 model
$ /usr/src/tensorrt/bin/trtexec --onnx=yolov7.onnx --saveEngine=yolov7fp16.engine --fp16
# int8 QAT model, the onnx model with Q&DQ nodes
$ /usr/src/tensorrt/bin/trtexec --onnx=yolov7qat.onnx --saveEngine=yolov7QAT.engine --fp16 --int8

Detection & Validate

• detect with image:

  $ ./build/detect --engine=yolov7db4fp32.engine --img=./imgs/horses.jpg,./imgs/zidane.jpg

• detect with video:
  • note: only support batchsize = 1 now.

  $ ./build/video_detect --engine=./yolov7fp32.engine --video=YOUR_VIDEO_PATH.mp4

• validate mAP on dataset
  • note: validate_coco only support model inputsize [batchsize, 3, 672, 672]

  $ ./build/validate_coco --engine=./yolov7fp32.engine --coco=/YOUR/COCO/DATA/PATH/
  --------------------------------------------------------
  Yolov7 initialized from: yolov7672.engine
  input : images , shape : [ 1,3,672,672,]
  output : output , shape : [ 1,27783,85,]
  --------------------------------------------------------
  5000 / 5000
  predict result has been written to ./predict.json
  
  $ python test_coco_map.py --predict ./predict.json --coco /YOUR/COCO/DATA/PATH/
  ...
  Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.51005
  ...