hylrh

> I have committed the model converted to OpenVINO IR, although I am still in the process of trial and error and haven't tested it yet. I hope I can...

you can use shufflenet/mobilenet as backbone to retrain this model ,or resize the input image to smaller size.after this work i get 20fps on mobile device

Hi i have found a c++ demo infer by ncnn, hope to help you. https://github.com/FeiGeChuanShu/yolov5-seg-ncnn