search

rishizek / tensorflow-deeplab-v3-plus

DeepLabv3+ built in TensorFlow
MIT License
833 stars 307 forks source link

About the evaluate.py #52

Open sunxiaochuanpr opened 5 years ago

sunxiaochuanpr commented 5 years ago

INFO:tensorflow:current step = 100 (27.043 sec) INFO:tensorflow:current step = 200 (18.439 sec) INFO:tensorflow:current step = 300 (16.081 sec) INFO:tensorflow:current step = 400 (16.369 sec) INFO:tensorflow:current step = 500 (17.911 sec) INFO:tensorflow:current step = 600 (15.671 sec) INFO:tensorflow:current step = 700 (16.864 sec) INFO:tensorflow:current step = 800 (15.698 sec) INFO:tensorflow:current step = 900 (14.314 sec) INFO:tensorflow:current step = 1000 (15.316 sec) INFO:tensorflow:current step = 1100 (14.541 sec) INFO:tensorflow:current step = 1200 (14.280 sec) INFO:tensorflow:current step = 1300 (13.763 sec) INFO:tensorflow:current step = 1400 (14.268 sec) Intersection over Union for each class: class 0: 0.9550 class 1: 0.0000 class 2: 0.0000 class 3: 0.0000 class 4: 0.0000 class 5: 0.0000 class 6: 0.0000 class 7: 0.0000 class 8: 0.0000 class 9: 0.0000 class 10: 0.0000 class 11: 0.0000 class 12: 0.0000 class 13: 0.0000 class 14: 0.0000 class 15: 0.0000 class 16: 0.0000 class 17: 0.0000 class 18: 0.0000 class 19: 0.0000 class 20: 0.0000 mean Intersection over Union: 0.0455 Pixel Accuracy: 0.9206

Process finished with exit code 0

why these classes are all 0.0000 except the class 0?

LiouCZ commented 5 years ago

你解决这个问题没有

LiouCZ commented 5 years ago

Have you solved the problem? I had the same problem

sunxiaochuanpr commented 5 years ago

你解决这个问题没有

应该是tfrecord没有制作好，重新create tfrecord，要用作者给的增强的数据集

LiouCZ commented 5 years ago

你解决这个问题没有

应该是tfrecord没有制作好，重新create tfrecord，要用作者给的增强的数据集

谢谢！可以留个邮箱交流一下吗？

LiouCZ commented 4 years ago

你的label数据集格式有问题，改成灰度图

------------------ 原始邮件 ------------------ 发件人: "GuoXiaoqing"notifications@github.com; 发送时间: 2019年8月10日(星期六) 晚上6:25 收件人: "rishizek/tensorflow-deeplab-v3-"tensorflow-deeplab-v3-plus@noreply.github.com; 抄送: "刘宸哲"595506991@qq.com; "Comment"comment@noreply.github.com; 主题: Re: [rishizek/tensorflow-deeplab-v3-plus] About the evaluate.py (#52)

你们怎么解决的这个问题呀？evaluate中没有用到tfrecord呀。

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

Guo-Xiaoqing commented 4 years ago

谢谢你，我已经解决了哈哈。

Loveling notifications@github.com于2019年8月10日 周六18:52写道：

你的label数据集格式有问题，改成灰度图

------------------ 原始邮件 ------------------ 发件人: "GuoXiaoqing"notifications@github.com; 发送时间: 2019年8月10日(星期六) 晚上6:25 收件人: "rishizek/tensorflow-deeplab-v3-"< tensorflow-deeplab-v3-plus@noreply.github.com>; 抄送: "刘宸哲"595506991@qq.com; "Comment"comment@noreply.github.com; 主题: Re: [rishizek/tensorflow-deeplab-v3-plus] About the evaluate.py (#52)

你们怎么解决的这个问题呀？evaluate中没有用到tfrecord呀。

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

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/rishizek/tensorflow-deeplab-v3-plus/issues/52?email_source=notifications&email_token=AIKMY2XX6AIQGLPNTHWYFA3QD2MYFA5CNFSM4HRQ4KW2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4ALEWI#issuecomment-520139353, or mute the thread https://github.com/notifications/unsubscribe-auth/AIKMY2QIFBN677LUQTND4L3QD2MYFANCNFSM4HRQ4KWQ .