The benchmark aims to check QR codes detection and decoding with classic OpenCV implementation from objdetect module and dnn based implementation in opencv_contrib repo contributed by the WeChat team.

To run the benchmark:

1. Download BoofCV dataset to path_folder. The dataset has been prepared by the community BoofCV: BoofCV main page.
2. Install opencv-python or build OpenCV.
3. Use "python qr.py -alg opencv -o out -p path_folder/qrcodes/detection" to run benchmark.

To run OpenCV and WeChat algorithm:

1. Follow the steps from the previous part.
2. Install opencv-contrib-python or build OpenCV with OpenCV contrib.
3. Download files detect.prototxt, detect.caffemodel, sr.prototxt, sr.caffemodel and set path_to_model.
4. Use "python qr.py -alg opencv_wechat -o out -p path_folder/qrcodes/detection" to run benchmark.

The benchmark will generate a distribution of errors for each category and save them to the "REPORT_YYYY_DD_MM" folder. The error is calculated by measuring the distance between the gold QR code and the nearest QR code using the metric specified in the "--metric" parameter, and this distance is saved in "log.txt". This file can be used to compare the results of different algorithms and identify areas for improvement.