This project used to create tensorrt yolov5 plugin which is adapted to ultralytics/yolov5 and references to https://github.com/wang-xinyu/tensorrtx.
The format of result is [box_num, xmin, xmax, class, score, ...].
You can use function Yolov5LayerPlugin::add_yolov5_layer() to add this plugin to your network. net_w and net_h equal input dimension w and h respectively. det0, det1 and det2 are outputs of onnx model exported from ultralytics/yolov5 model without detect mode.
Some other functions, such as ByteStreamReader which is used to do serialization and deserialization.