rpn_loc_loss is nan, fast_rcnn_loc_loss is 0.0

[Llq201809]

I have changed the database and converted to tfrecords . when i run it for minutes, fast_rcnn_loc_loss become 0.0 rpn_loc_loss become nan.

``

for t_w = np.log(w/reference_w), I have check picture ,w >0 .So I think reference_w is wrong, I don't know how to modify it .Ask for your help.

[yangxue0827]

I speculate that it may be a problem with the data set. You can check if the data read from tfrecord is correct. @Llq201809

[Llq201809]

Thanks for your reply very much. I find error in my data set. Now it works . @yangxue0827

[Rachquazar]

@Llq201809 in my case, fast_rcnn_loc_loss=0.0. I am using a custom dataset. How did you find the error in your dataset? My dataset looks fine.

[Llq201809]

I have writed a print function to delete error coordinate when translate dataset to VOC format. For example,if Xmin<Xmax or Ymin<Ymax，print image then delete it manually. ------------------ 原始邮件 ------------------ 发件人: "Rachquazar"[email protected] 发送时间: 2019年8月30日(星期五) 下午5:57 收件人: "DetectionTeamUCAS/Cascade-RCNN_Tensorflow"[email protected]; 抄送: "Llq201809"[email protected];"Mention"[email protected]; 主题: Re: [DetectionTeamUCAS/Cascade-RCNN_Tensorflow] rpn_loc_loss is nan,fast_rcnn_loc_loss is 0.0 (#8)

@Llq201809 in my case, fast_rcnn_loc_loss=0.0. I am using a custom dataset. How did you find the error in your dataset? My dataset looks fine.

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

[Llq201809]

Sorry，it should be Xmax-Xmin<=0 or Ymax-Ymin>=0，delete it. ------------------ 原始邮件 ------------------ 发件人: "Rachquazar"[email protected] 发送时间: 2019年8月30日(星期五) 下午5:57 收件人: "DetectionTeamUCAS/Cascade-RCNN_Tensorflow"[email protected]; 抄送: "Llq201809"[email protected];"Mention"[email protected]; 主题: Re: [DetectionTeamUCAS/Cascade-RCNN_Tensorflow] rpn_loc_loss is nan,fast_rcnn_loc_loss is 0.0 (#8)

@Llq201809 in my case, fast_rcnn_loc_loss=0.0. I am using a custom dataset. How did you find the error in your dataset? My dataset looks fine.

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