ncnn-android-yolov5 icon indicating copy to clipboard operation
ncnn-android-yolov5 copied to clipboard

The YOLOv5 object detection android example

Results 46 ncnn-android-yolov5 issues
Sort by recently updated
recently updated
newest added

Your project only works on images. How do you make it work on a continuous video stream coming from a camera?

![21b492dc83382a08f85fb9c417608af](https://user-images.githubusercontent.com/65895569/103861682-c3155a00-50f8-11eb-9654-aa56e0910d3b.jpg)

为了提高识别效率我更换了v5n和v5n-lite-e测试发现精度会大幅度降低。 这种情况我想要提高精度并且保持一个不错的识别速度改怎么做

您好@nihui 我使用yolov5生成一个best.pt,然后使用命令python export.py --weights best.pt --int8 --include onnx --simplify导出onnx模型,再通过pnnx best.onnx inputshape=[1,3,640,640]生成了best.ncnn.bin和best.ncnn.param文件,替换这个工程的文件,点击识别没任何反应日志输出如下: 35.89ms detect 2024-10-10 18:20:29.261 6891-6905 OpenGLRenderer com.tencent.yolov5ncnn I Davey! duration=9223260857263ms; Flags=0, FrameTimelineVsyncId=392958, IntendedVsync=111179591303092, Vsync=111179641303090, InputEventId=801281897, HandleInputStart=111179645622482, AnimationStart=111179645675451, PerformTraversalsStart=111179646356805, DrawStart=111179646678836,...