Stephen Zhao
Stephen Zhao
> used your training code, the accuracy is about 50%，but the paper's accuracy has achieved 90%,could you please tell me why? I have met the similar question.On my dataset,the accuracy...
直接标线段，生成mask就行，自己写一下dataset方法，inputs就是原图，target就是生成的mask
> ### Search before asking > * [x] I have searched the YOLOv5 [issues](https://github.com/ultralytics/yolov5/issues) and [discussions](https://github.com/ultralytics/yolov5/discussions) and found no similar questions. > > ### Question > I am using following...
自己写一个dataset方法，核心就是训练数据是原图，标签是mask，一一对应
you can rewrite the pytorch dataset class, input is raw image and gt is mask png
thx for your quick reply, appreciate for your great job.\ my commant is as follow, fmo-desktop --input /home/zhaoqihui/Repo/fmo-cpp-demo-master/videos/770011352299781.mp4 --demo --record-dir /home/zhaoqihui/Repo/fmo-cpp-demo-master\ I get the error\ OpenCV(3.4.20-dev) /home/zhaoqihui/Repo/opencv-3.4/modules/videoio/src/cap_images.cpp:260: error: (-215:Assertion failed)...
> 我使用box2mask_r50_lsj_8x2_50e_duanmian.py配置文件训练模型，模型训练完成后导出onnx报错，在debug过程中发现会跳转到maskformer.py中，但是跳转到其中的self.onnx_export()函数时，函数为空，raise NotImplementedError。 请问是我使用错误，还是代码并未实现呢？single_stage.py中是有onnx_export.py实现的 请问你转成功了吗？我们这边也需要做部署，做tensorrt或者openvino的
我也是同样的错误，是不是这个只能支持那种000001.jpg这样子名称的数据集
any update? I have met the same problem.
looking forward to your reply,thx a lot.