Test with trained network but return null detection results

[QianGuo123]

I have trained a network and the result shown in Tensorboard is pretty good. However, when I test the trained network with test dataset, even train dataset, there is no detection bbox.

Did someone ever meet this situation and what should I do?

Thanks very much!

[liuqianqian77]

Hello, do you use your own data set for training?

[QianGuo123]

Yes, I trained it with my own data whose format is same to DOTA dataset. The train process in Tensorboard is pretty good but it just doesn't work when testing. I think there may be some problem in the test_dota.py but I can't figure it out....

[liuqianqian77]

您好，我用中文和您交流吧，我目前遇到了一个问题想请教您一下，我也是用自己的数据集进行训练的，在将数据转tfrocord的时候，我遇到了KeyError: 'eddy\n这个问题，但是一些设置我已经按照作者的进行修改了。期待您的答复，非常感谢！

…

------------------&nbsp;原始邮件&nbsp;------------------ 发件人: "DetectionTeamUCAS/RetinaNet_Tensorflow_Rotation" <notifications@github.com&gt;; 发送时间:&nbsp;2020年7月30日(星期四) 上午10:47 收件人:&nbsp;"DetectionTeamUCAS/RetinaNet_Tensorflow_Rotation"<RetinaNet_Tensorflow_Rotation@noreply.github.com&gt;; 抄送:&nbsp;"刘倩倩"<2434070972@qq.com&gt;;"Comment"<comment@noreply.github.com&gt;; 主题:&nbsp;Re: [DetectionTeamUCAS/RetinaNet_Tensorflow_Rotation] Test with trained network but return null detection results (#72) Yes, I trained it with my own data whose format is same to DOTA dataset. The train process in Tensorboard is pretty good but it just doesn't work when testing. I think there may be some problem in the test_dota.py but I can't figure it out.... | | 郭倩 | | 15608225070@163.com | 签名由网易邮箱大师定制 On 7/30/2020 10:42，liuqianqian77<notifications@github.com&gt; wrote： Hello, do you use your own data set for training? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe. — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

[QianGuo123]

我没有遇到你这个问题呢，你是用的DOTA格式的数据集吗？ 或许你可以尝试先把数据转成DOTA 格式的然后再跑

[liuqianqian77]

恩恩，对的。我看到作者的这个代码里边是将DOTA数据集转成VOC样式的，我是将自己的数据集直接转成VOC样式的了，谢谢您的答复，我再搜索一下找解决方法吧。

…

------------------&nbsp;原始邮件&nbsp;------------------ 发件人: "DetectionTeamUCAS/RetinaNet_Tensorflow_Rotation" <notifications@github.com&gt;; 发送时间:&nbsp;2020年7月30日(星期四) 中午11:00 收件人:&nbsp;"DetectionTeamUCAS/RetinaNet_Tensorflow_Rotation"<RetinaNet_Tensorflow_Rotation@noreply.github.com&gt;; 抄送:&nbsp;"刘倩倩"<2434070972@qq.com&gt;;"Comment"<comment@noreply.github.com&gt;; 主题:&nbsp;Re: [DetectionTeamUCAS/RetinaNet_Tensorflow_Rotation] Test with trained network but return null detection results (#72) 我没有遇到你这个问题呢，你是用的DOTA格式的数据集吗？ 或许你可以尝试先把数据转成DOTA 格式的然后再跑 — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

[QianGuo123]

啊原来是在test_dota.py的文件中有一个关于网络名称的设置忘改了，修改后问题就解决了