Skip to content

Latest commit

 

History

History
196 lines (121 loc) · 5.67 KB

README.md

File metadata and controls

196 lines (121 loc) · 5.67 KB

yolov7

Table of contents

1. Description

The model used in this example comes from the following open source projects:

https://github.com/airockchip/yolov7

2. Current Support Platform

RK3566, RK3568, RK3588, RK3562, RK1808, RV1109, RV1126

3. Pretrained Model

Download link:

./yolov7-tiny.onnx
./yolov7.onnx

Download with shell command:

cd model
./download_model.sh

Note: The model provided here is an optimized model, which is different from the official original model. Take yolov7-tiny.onnx as an example to show the difference between them.

  1. The comparison of their output information is as follows. The left is the official original model, and the right is the optimized model. The three colored boxes in the figure represent the changes of the three outputs.
Image
  1. Taking the output change [1,3,20,20,85]->[1,255,20,20] as an example, we remove the subgraphs behind the convolution node in the model (the framed part in the figure), and keep the output of the convolution ([1,255,20,20]).
Image

4. Convert to RKNN

Usage:

cd python
python convert.py <onnx_model> <TARGET_PLATFORM> <dtype(optional)> <output_rknn_path(optional)>

# such as: 
python convert.py ../model/yolov7-tiny.onnx rk3588
# output model will be saved as ../model/yolov7.rknn

Description:

  • <onnx_model>: Specify ONNX model path.
  • <TARGET_PLATFORM>: Specify NPU platform name. Support Platform refer [here](#2 Current Support Platform).
  • <dtype>(optional): Specify as i8, u8 or fp. i8/u8 for doing quantization, fp for no quantization. Default is i8.
  • <output_rknn_path>(optional): Specify save path for the RKNN model, default save in the same directory as ONNX model with name yolov7.rknn

5. Python Demo

Usage:

cd python
# Inference with PyTorch model or ONNX model
python yolov7.py --model_path <pt_model/onnx_model> --img_show

# Inference with RKNN model
python yolov7.py --model_path <rknn_model> --target <TARGET_PLATFORM> --img_show

Description:

  • <TARGET_PLATFORM>: Specify NPU platform name. Support Platform refer [here](#2 Current Support Platform).

  • <pt_model / onnx_model / rknn_model>: Specify the model path.

6. Android Demo

Note: RK1808, RV1109, RV1126 does not support Android.

6.1 Compile and Build

Please refer to the Compilation_Environment_Setup_Guide document to setup a cross-compilation environment and complete the compilation of C/C++ Demo.
Note: Please replace the model name with yolov7.

6.2 Push demo files to device

With device connected via USB port, push demo files to devices:

adb root
adb remount
adb push install/<TARGET_PLATFORM>_android_<ARCH>/rknn_yolov7_demo/ /data/

6.3 Run demo

adb shell
cd /data/rknn_yolov7_demo

export LD_LIBRARY_PATH=./lib
./rknn_yolov7_demo model/yolov7.rknn model/bus.jpg

  • After running, the result was saved as out.png. To check the result on host PC, pull back result referring to the following command:

    adb pull /data/rknn_yolov7_demo/out.png

  • Output result refer Expected Results.

7. Linux Demo

7.1 Compile and Build

Please refer to the Compilation_Environment_Setup_Guide document to setup a cross-compilation environment and complete the compilation of C/C++ Demo.
Note: Please replace the model name with yolov7.

7.2 Push demo files to device

  • If device connected via USB port, push demo files to devices:
adb push install/<TARGET_PLATFORM>_linux_<ARCH>/rknn_yolov7_demo/ /userdata/
  • For other boards, use scp or other approaches to push all files under install/<TARGET_PLATFORM>_linux_<ARCH>/rknn_yolov7_demo/ to userdata.

7.3 Run demo

adb shell
cd /userdata/rknn_yolov7_demo

export LD_LIBRARY_PATH=./lib
./rknn_yolov7_demo model/yolov7.rknn model/bus.jpg

  • After running, the result was saved as out.png. To check the result on host PC, pull back result referring to the following command:

    adb pull /userdata/rknn_yolov7_demo/out.png

  • Output result refer Expected Results.

8. Expected Results

This example will print the labels and corresponding scores of the test image detect results, as follows:

person @ (212 241 285 511) 0.886
bus @ (86 134 540 444) 0.855
person @ (476 237 561 519) 0.835
person @ (112 234 218 531) 0.835
person @ (79 330 124 524) 0.346

  • Note: Different platforms, different versions of tools and drivers may have slightly different results.