search

danfenghong / IEEE_TGRS_MDL-RS

Danfeng Hong, Lianru Gao, Naoto Yokoya, Jing Yao, Jocelyn Chanussot, Qian Du, Bing Zhang. More Diverse Means Better: Multimodal Deep Learning Meets Remote Sensing Imagery Classification, IEEE TGRS, 2021, 59(5): 4340-4354.
102 stars 21 forks source link

关于Early_fusion_CNNs.py的训练结果 #4

Open 4fee8fea opened 2 years ago

4fee8fea commented 2 years ago

您好！

    我直接运行了Early_fusion_CNNs.py这一脚本，得到的运行结果如下：

Screenshot from 2021-11-13 18-31-01

感觉报告的准确率出了点问题，请问这是正常的吗？

谢谢！

danfenghong commented 2 years ago

应该是数据导入出问题了。

Di Xiu @.***> 于2021年11月13日周六 下午6:32写道：

您好！

我直接运行了`Early_fusion_CNNs.py`，得到的运行结果如下：

[image: Screenshot from 2021-11-13 18-31-01] https://user-images.githubusercontent.com/53552367/141615261-7cd678dc-2ce4-471e-bea8-5ed95c1bad10.png 感觉报告的准确率出了点问题，请问这是正常的吗？

谢谢！

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/danfenghong/IEEE_TGRS_MDL-RS/issues/4, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFL2GZROKHOPIX7EXSNKDLTULY5CPANCNFSM5H6PSYBA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

4fee8fea commented 2 years ago

感谢您的迅速答复！

我注意到Early_fusion_CNNs.py中MODEL初始指定为'CML-HSI'。我将其改为'MML'后重新进行了一次训练，结果如下：

Early_fusion_CNNs_result_MML

仍然是一个准确率较低的结果。

Note that HSI_TrSet.mat and HSI_TeSet.mat need to download manually from the given the links of google drive or baiduyun, due to its large size. Google drive: https://drive.google.com/file/d/1vno8vQxCXgr7xk-Nez9CVbJmjkDMBCbe/view?usp=sharing Baiduyun: https://pan.baidu.com/s/1ug_tKyrbwg_CzHGpB2YK2A (access code: hfw3)

我在Google drive中下载了数据集，将HSI_TeSet.mat和HSI_TrSet.mat放入HSI_LiDAR_CNN文件夹中，源码部分不做修改，得到如上结果。

请问数据导入部分需要做额外处理吗？

谢谢！

danfenghong commented 2 years ago

你可能需要去关注下你使用的是多模态还是交叉模态

Di Xiu @.***> 于2021年11月13日周六 下午7:17写道：

感谢您的迅速答复！

我注意到Early_fusion_CNNs.py中MODEL初始指定为'CML-HSI'。我将其改为'MML'后重新进行了一次训练，结果如下：

[image: Early_fusion_CNNs_result_MML] https://user-images.githubusercontent.com/53552367/141634023-0072bf16-0ab3-48fd-a3da-2e2c93d2123c.png

仍然是一个准确率较低的结果。

Note that HSI_TrSet.mat and HSI_TeSet.mat need to download manually from the given the links of google drive or baiduyun, due to its large size. Google drive: https://drive.google.com/file/d/1vno8vQxCXgr7xk-Nez9CVbJmjkDMBCbe/view?usp=sharing Baiduyun: https://pan.baidu.com/s/1ug_tKyrbwg_CzHGpB2YK2A (access code: hfw3)

我在Google drive中下载了数据集，将HSI_TeSet.mat和HSI_TrSet.mat放入HSI_LiDAR_CNN文件夹中，源码部分不做修改，得到如上结果。

请问数据导入部分需要做额外处理吗？

谢谢！

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/danfenghong/IEEE_TGRS_MDL-RS/issues/4#issuecomment-968052306, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFL2GZU3QHDWRZ2GN5RJLJDULZCNBANCNFSM5H6PSYBA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

4fee8fea commented 2 years ago

谢谢您答复！

上面两次训练中，我在MDL-RS_CNNs/Early_fusion_CNNs.py中变更的代码只有line 271这一处：

Early_fusion_CNNs_MODEL_setting

上面这两行分别对应了多模态与交叉模态，但是训练结果都有一点不对劲

suosuoxi commented 2 years ago

跨越时空的回答hhh，可能是python的版本不一致，我用3.6的时候遇到一模一样的情况，改成3.7后就都正常。