此demo演示了caffe框架下的yolov3模型在MLU270上的移植流程。
本示例基于 Neuware 1.6.1 + Python2.7 版本测试通过。
相关教程链接:https://developer.cambricon.com/index/curriculum/expdetails/id/4/classid/8.html
bash setup.sh
python darknet2caffe-yoloV23.py 3 yolov3.cfg yolov3.weights yolov3.prototxt yolov3.caffemodel
在转换为prototxt文件后,还需要添加后处理节点,该步骤请详细参考教程:https://developer.cambricon.com/index/curriculum/expdetails/id/4/classid/8.html
./generate_quantized_pt --ini_file convert_quantized.ini
python detect.py yolov3_int8.prototxt yolov3.caffemodel
./caffe genoff -model ../quantize/yolov3_int8.prototxt -weights ../quantize/yolov3.caffemodel -mcore MLU270
编译:bash make.sh
运行:bash run.sh