How to train from scratch? Dataset location

[Russzheng]

Hi, sorry silly question, I am a newbie to COCO and openPose, I am just wondering where should I put all the datasets/mask/json and what directory should I pass to train the shufflenet version?

Right now, loading dataset giving me shell train dataset len: 121522 val dataset len: 4873

But shell cv2.error: OpenCV(4.0.0) /io/opencv/modules/imgproc/src/resize.cpp:3784: error: (-215:Assertion failed) !ssize.empty() in function 'resize'

[GreenTeaHua]

the same question, do u got the answer?

[xdl932365660]

the same question,I want a anwser,please!!

[xdl932365660]

@Risszheng

[tensorboy]

The bug comes from the path of the mask:

https://github.com/tensorboy/pytorch_Realtime_Multi-Person_Pose_Estimation/blob/master/training/datasets/coco_data/COCO_data_pipeline.py#L226-L231

And I will fix it now!

[Russzheng]

The bug comes from the path of the mask:

https://github.com/tensorboy/pytorch_Realtime_Multi-Person_Pose_Estimation/blob/master/training/datasets/coco_data/COCO_data_pipeline.py#L226-L231

And I will fix it now!

Thanks so much!

[tensorboy]

The bug is come from the path of the mask: https://github.com/tensorboy/pytorch_Realtime_Multi-Person_Pose_Estimation/blob/master/training/datasets/coco_data/COCO_data_pipeline.py#L226-L231

I will fix it now!

[xdl932365660]

Thanks for your help！Would you like to try to convert the pytorch modle to caffemodel ?

------------------ 原始邮件 ------------------ 发件人: "tensorboy"[email protected]; 发送时间: 2019年4月7日(星期天) 上午9:04 收件人: "tensorboy/pytorch_Realtime_Multi-Person_Pose_Estimation"[email protected]; 抄送: "传奇"[email protected]; "Comment"[email protected]; 主题: Re: [tensorboy/pytorch_Realtime_Multi-Person_Pose_Estimation] How totrain from scratch? Dataset location (#92)

The bug is come from the path of the mask: https://github.com/tensorboy/pytorch_Realtime_Multi-Person_Pose_Estimation/blob/master/training/datasets/coco_data/COCO_data_pipeline.py#L226-L231

I will fix it now!

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.