I ran out of this code, but the accuracy is too low.

IcedDoggie / Micro-Expression-with-Deep-Learning

Experimentation of deep learning on the subjects of micro-expression spotting and recognition.

267 stars 102 forks source link

I ran out of this code, but the accuracy is too low. #20

Open yuanboya opened 5 years ago

yuanboya commented 5 years ago

 Hi, I ran the example for single db,the parameter is batch_size=1,epoch=70,patience=20,but the accuracy is too low,is only 0.34,it have large different from your results.Can you tell me how to solve this problem?Thank you.
Looking forward to you reply!

IcedDoggie commented 5 years ago

The F1 for single database(CASME II) is 0.5 approximately and 0.5244 for accuracy and 0.4396 for UAR, with CASME II Optical (optical flow image) and Optical Strain using Temporal Enrichment. Maybe the data parsed in is incorrct?

ZYDi commented 5 years ago

 Hi, I ran the example for single db,the parameter is batch_size=1,epoch=70,patience=20,but the accuracy is too low,is only 0.34,it have large different from your results.Can you tell me how to solve this problem?Thank you.
Looking forward to you reply!

how to run TIM?

IcedDoggie commented 5 years ago

TIM please refer to original author's website, i believe there's code provided there.

ZYDi commented 5 years ago

TIM please refer to original author's website, i believe there's code provided there.

I download it ，but i don't know how to run it.

weijieZhu0803 commented 5 years ago

 Hi, I ran the example for single db,the parameter is batch_size=1,epoch=70,patience=20,but the accuracy is too low,is only 0.34,it have large different from your results.Can you tell me how to solve this problem?Thank you.
Looking forward to you reply!

 Hi, I ran the example for single db,the parameter is batch_size=1,epoch=70,patience=20,but the accuracy is too low,is only 0.34,it have large different from your results.Can you tell me how to solve this problem?Thank you.
Looking forward to you reply!

For this paper,when I run those code ,I meeting some difficulties. I want to talk to you about ME recognition.Can add you wechat or QQ?

yuanboya commented 5 years ago

 Hi, I ran the example for single db,the parameter is batch_size=1,epoch=70,patience=20,but the accuracy is too low,is only 0.34,it have large different from your results.Can you tell me how to solve this problem?Thank you.
Looking forward to you reply!

 Hi, I ran the example for single db,the parameter is batch_size=1,epoch=70,patience=20,but the accuracy is too low,is only 0.34,it have large different from your results.Can you tell me how to solve this problem?Thank you.
Looking forward to you reply!

For this paper,when I run those code ,I meeting some difficulties. I want to talk to you about ME recognition.Can add you wechat or QQ?

qq 649685276

leohwang commented 4 years ago

The F1 for single database(CASME II) is 0.5 approximately and 0.5244 for accuracy and 0.4396 for UAR, with CASME II Optical (optical flow image) and Optical Strain using Temporal Enrichment. Maybe the data parsed in is incorrct?

Hi, I have read your FG paper "Enriched long-term recurrent convolutional network for facial micro-expression recognition" in FG 2019. But I tried to reproduce the results of ELRCN-SE and ELRCN-TE in the Table 1. Unfortunately, I obtained the different results that ELRCN-SE outperforms ELRCN-TE. I also see that you replied other people here about the method and result. However, in Section III. C and figure 1, both of ELRCN-SE and ELRCN-TE should contain gray-level, optical flow and optical strain for micro-expression recognition. It caused me very confused about your paper and your reply.

So I would like to confirm what kind of information for ELRCN-SE for Table 1 in your FG paper? Only using optical flow and optical strain for temporal enrichment? Thanks very much.

Xu-Fengnian commented 3 years ago

Hello, I am currently trying to run the program, but I don’t know where to start. Can you leave your contact information? Or please add my QQ987159382. thank you very much

IcedDoggie commented 3 years ago

Hi Fengnian,

Sorry about the messy codes. Have you tried reading the documentation below? there is a section about trying to run from scratch.

Best Regards, Huai-Qian Khor -- Journey of thousand miles begins with a single step, Doing the best at this moment puts you in the best place for the next moment.

On Sat, 10 Apr 2021 at 19:46, Xu-Fengnian @.***> wrote:

Hello, I am currently trying to run the program, but I don’t know where to start. Can you leave your contact information? Or please add my QQ987159382. thank you very much

— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub https://github.com/IcedDoggie/Micro-Expression-with-Deep-Learning/issues/20#issuecomment-817123815, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA7ZNEHPO5KVRXDVZTR64FTTIA3ANANCNFSM4GI76X4A .

Xu-Fengnian commented 3 years ago

嗨，丰年，对不起，代码混乱。您是否尝试阅读下面的文档？有一节关于尝试从头开始运行。最好的问候，淮干·霍尔（Huai-Qian Khor）-千里旅程始于一步，此刻尽力而为，将您带入下一时刻的最佳去处。 … 在2021年4月10日，星期六，19：46，Xu-Fengnian @.***>写道：您好，我目前正在尝试运行该程序，但是我不知道从哪里开始。您可以留下您的联系信息吗？或者，请添加我的QQ987159382。非常感谢您—之所以收到此信息，是因为您修改了打开/关闭状态。直接回复此电子邮件，在GitHub < ＃20（评论） >上查看，或退订< https://github.com/notifications/unsubscribe-auth/AA7ZNEHPO5KVRXDVZTR64FTTIA3ANANCNFSM4GI76X4A >。

Thank you very much for your reply. I am currently trying out the TIM code. How do I set the length of the TIM? I saw that you updated the link to download the program, but the link seems to be no longer working. Can you send the TIM code to my email address yangxinrui1028@gmail.com?

IcedDoggie commented 3 years ago

Hello,

I need some time to dig back into my own files. in the meantime, how about try this first ? https://github.com/zsdust/TIM-temporal-interpolation-model

I am not sure if it works.

Best Regards, Huai-Qian Khor -- Journey of thousand miles begins with a single step, Doing the best at this moment puts you in the best place for the next moment.

On Tue, 13 Apr 2021 at 18:22, Xu-Fengnian @.***> wrote:

嗨，丰年，对不起，代码混乱。您是否尝试阅读下面的文档？有一节关于尝试从头开始运行。最好的问候，淮干·霍尔（Huai-Qian Khor）- 千里旅程始于一步，此刻尽力而为，将您带入下一时刻的最佳去处。 … <#m-1213837524030853595> 在2021年4月10日，星期六，19：46，Xu-Fengnian @.***>写道：您好，我目前正在尝试运行该程序，但是我不知道从哪里开始。您可以留下您的联系信息吗？或者，请添加我的QQ987159382。非常感谢您—之所以收到此信息，是因为您修改了打开/关闭状态。直接回复此电子邮件，在GitHub < ＃20（评论） https://github.com/IcedDoggie/Micro-Expression-with-Deep-Learning/issues/20#issuecomment-817123815

上查看，或退订< https://github.com/notifications/unsubscribe-auth/AA7ZNEHPO5KVRXDVZTR64FTTIA3ANANCNFSM4GI76X4A 。

Thank you very much for your reply. I am currently trying out the TIM code. How do I set the length of the TIM? I saw that you updated the link to download the program, but the link seems to be no longer working. Can you send the TIM code to my email address @.***?

— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub https://github.com/IcedDoggie/Micro-Expression-with-Deep-Learning/issues/20#issuecomment-818626612, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA7ZNEENP4XFIVYGAJFLWSTTIQLO3ANCNFSM4GI76X4A .

Xu-Fengnian commented 3 years ago

Hello, I need some time to dig back into my own files. in the meantime, how about try this first ? https://github.com/zsdust/TIM-temporal-interpolation-model I am not sure if it works. Best Regards, Huai-Qian Khor -- Journey of thousand miles begins with a single step, Doing the best at this moment puts you in the best place for the next moment. … On Tue, 13 Apr 2021 at 18:22, Xu-Fengnian @.> wrote: 嗨，丰年，对不起，代码混乱。您是否尝试阅读下面的文档？有一节关于尝试从头开始运行。最好的问候，淮干·霍尔（Huai-Qian Khor）- 千里旅程始于一步，此刻尽力而为，将您带入下一时刻的最佳去处。 … <#m-1213837524030853595> 在2021年4月10日，星期六，19：46，Xu-Fengnian @.>写道：您好，我目前正在尝试运行该程序，但是我不知道从哪里开始。您可以留下您的联系信息吗？或者，请添加我的QQ987159382。非常感谢您—之所以收到此信息，是因为您修改了打开/关闭状态。直接回复此电子邮件，在GitHub < ＃20（评论） <#20 (comment)> >上查看，或退订< https://github.com/notifications/unsubscribe-auth/AA7ZNEHPO5KVRXDVZTR64FTTIA3ANANCNFSM4GI76X4A >。 Thank you very much for your reply. I am currently trying out the TIM code. How do I set the length of the TIM? I saw that you updated the link to download the program, but the link seems to be no longer working. Can you send the TIM code to my email address @.***? — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#20 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA7ZNEENP4XFIVYGAJFLWSTTIQLO3ANCNFSM4GI76X4A .

Thank you very much for your help. I downloaded the program through the link you provided. After running it, I found that the length of the video did change. But I want to know, does TIM10 mean to modify the length of the video to 10s? After I run the program repeatedly, the video length is always 2900fps and the time is 14s. Is this result correct? If it is correct, after running the TIM10 program, do I need to convert the video to each frame of pictures, and then select 9 of them as the data in CASME2_Optical?

IcedDoggie commented 3 years ago

HI FengNian,

yes modify it from N frames to 10 frames. yes after running then perform optical flow extraction. Sorry I have not searched my files, will do so promptly

Best Regards, Huai-Qian Khor -- Journey of thousand miles begins with a single step, Doing the best at this moment puts you in the best place for the next moment.

On Thu, 15 Apr 2021 at 13:23, Xu-Fengnian @.***> wrote:

Hello, I need some time to dig back into my own files. in the meantime, how about try this first ? https://github.com/zsdust/TIM-temporal-interpolation-model I am not sure if it works. Best Regards, Huai-Qian Khor --

Journey of thousand miles begins with a single step, Doing the best at this moment puts you in the best place for the next moment. … <#m-7578797441964957885> On Tue, 13 Apr 2021 at 18:22, Xu-Fengnian @.> wrote: 嗨，丰年，对不起，代码混乱。您是否尝试阅读下面的文档？有一节关于尝试从头开始运行。最好的问候，淮干·霍尔（Huai-Qian Khor）- 千里旅程始于一步，此刻尽力而为，将您带入下一时刻的最佳去处。 … <#m-1213837524030853595> 在2021年4月10日，星期六，19：46，Xu-Fengnian @.>写道：您好，我目前正在尝试运行该程序，但是我不知道从哪里开始。您可以留下您的联系信息吗？或者，请添加我的QQ987159382。非常感谢您—之所以收到此信息，是因为您修改了打开/关闭状态。直接回复此电子邮件，在GitHub < ＃20（评论） <#20 (comment) https://github.com/IcedDoggie/Micro-Expression-with-Deep-Learning/issues/20#issuecomment-817123815>

上查看，或退订< https://github.com/notifications/unsubscribe-auth/AA7ZNEHPO5KVRXDVZTR64FTTIA3ANANCNFSM4GI76X4A https://github.com/notifications/unsubscribe-auth/AA7ZNEHPO5KVRXDVZTR64FTTIA3ANANCNFSM4GI76X4A 。 Thank you very much for your reply. I am currently trying out the TIM code. How do I set the length of the TIM? I saw that you updated the link to download the program, but the link seems to be no longer working. Can you send the TIM code to my email address @.***? — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#20 (comment) https://github.com/IcedDoggie/Micro-Expression-with-Deep-Learning/issues/20#issuecomment-818626612>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA7ZNEENP4XFIVYGAJFLWSTTIQLO3ANCNFSM4GI76X4A .

Thank you very much for your help. I downloaded the program through the link you provided. After running it, I found that the length of the video did change. But I want to know, does TIM10 mean to modify the length of the video to 10s? After I run the program repeatedly, the video length is always 2900fps and the time is 14s. Is this result correct? If it is correct, after running the TIM10 program, do I need to convert the video to each frame of pictures, and then select 9 of them as the data in CASME2_Optical?

— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub https://github.com/IcedDoggie/Micro-Expression-with-Deep-Learning/issues/20#issuecomment-820114971, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA7ZNEGQAZ4O5GF5CFMHEB3TIZZ5DANCNFSM4GI76X4A .

Xu-Fengnian commented 3 years ago

HI FengNian，是的，将其从N帧修改为10帧。是，运行后再执行光流提取。抱歉，我还没有搜索我的文件，会尽快这样做。最好的问候，Huai-Qian Khor-千里之行始于一步，此刻尽力而为，将使您在下一刻成为最好的地方。 … On Thu, 15 Apr 2021 at 13:23, Xu-Fengnian @.> wrote: Hello, I need some time to dig back into my own files. in the meantime, how about try this first ? https://github.com/zsdust/TIM-temporal-interpolation-model I am not sure if it works. Best Regards, Huai-Qian Khor -- Journey of thousand miles begins with a single step, Doing the best at this moment puts you in the best place for the next moment. … <#m-7578797441964957885> On Tue, 13 Apr 2021 at 18:22, Xu-Fengnian @.> wrote: 嗨，丰年，对不起，代码混乱。您是否尝试阅读下面的文档？有一节关于尝试从头开始运行。最好的问候，淮干·霍尔（Huai-Qian Khor）- 千里旅程始于一步，此刻尽力而为，将您带入下一时刻的最佳去处。 … <#m-1213837524030853595> 在2021年4月10日，星期六，19：46，Xu-Fengnian @.>写道：您好，我目前正在尝试运行该程序，但是我不知道从哪里开始。您可以留下您的联系信息吗？或者，请添加我的QQ987159382。非常感谢您—之所以收到此信息，是因为您修改了打开/关闭状态。直接回复此电子邮件，在GitHub < ＃20（评论） <#20 (comment) <#20 (comment)>> >上查看，或退订< https://github.com/notifications/unsubscribe-auth/AA7ZNEHPO5KVRXDVZTR64FTTIA3ANANCNFSM4GI76X4A https://github.com/notifications/unsubscribe-auth/AA7ZNEHPO5KVRXDVZTR64FTTIA3ANANCNFSM4GI76X4A >。 Thank you very much for your reply. I am currently trying out the TIM code. How do I set the length of the TIM? I saw that you updated the link to download the program, but the link seems to be no longer working. Can you send the TIM code to my email address @.? — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#20 (comment) <#20 (comment)>>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA7ZNEENP4XFIVYGAJFLWSTTIQLO3ANCNFSM4GI76X4A . Thank you very much for your help. I downloaded the program through the link you provided. After running it, I found that the length of the video did change. But I want to know, does TIM10 mean to modify the length of the video to 10s? After I run the program repeatedly, the video length is always 2900fps and the time is 14s. Is this result correct? If it is correct, after running the TIM10 program, do I need to convert the video to each frame of pictures, and then select 9 of them as the data in CASME2_Optical? — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#20 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA7ZNEGQAZ4O5GF5CFMHEB3TIZZ5DANCNFSM4GI76X4A .

Thank you very much. Now that I have a VGG model, how can I evaluate it?