Open sudonto opened 2 years ago
I tried to train on the ucf101 dataset, but f1 was about 84%. I have the same problem as you.Have you solved the problem.-- I do not use the tune_from=./checkpoint/TEA_something_RGB_tea50_shift8_blockres_avg_segment8_e50/ckpt.best.pth.tar . I do not have checkpoint of TEA trained on Kinetics 400 dataset. If you get it ,would you mind to send me ?
I have the same problem as well. I need the checkpoint of TEA trained on K-400 dataset to pretrain my model.
Hello, would you share your trained model of TEA trained on Somethin v1 ? I need the checkpoint of TEA to pretrain my model, but it`s sooo hard for me to train TEA on Somethin v1 with not enough computing sources. It would be appreciated if you could share.
Hello, I gotta check weather I save the trained model you want. If I find it I will send it to you in a few days. I am willing to help you for your research. If I can't find it, maybe I will try to train the same model on sthv1 for you.
---- Replied Message ---- | From | @.> | | Date | 06/08/2023 17:40 | | To | @.> | | Cc | @.>@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) |
Hello, would you share your trained model of TEA trained on Somethin v1 ? I need the checkpoint of TEA to pretrain my model, but it`s sooo hard for me to train TEA on Somethin v1 with not enough computing sources. It would be appreciated if you could share.
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>
Hello, I gotta check weather I save the trained model you want. If I find it I will send it to you in a few days. I am willing to help you for your research. If I can't find it, maybe I will try to train the same model on sthv1 for you. … ---- Replied Message ---- | From | @.> | | Date | 06/08/2023 17:40 | | To | @.> | | Cc | @.>@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) | Hello, would you share your trained model of TEA trained on Somethin v1 ? I need the checkpoint of TEA to pretrain my model, but it`s sooo hard for me to train TEA on Somethin v1 with not enough computing sources. It would be appreciated if you could share. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>
Oh, my God. I have never dreamed about your reply so quickly. Apologize for my late reply. Appreciate your willingness to help me with my research. Specifically, I used tea50_8f in my experiment, so I think I just need the model trained by /scrips/train_tea_something_rgb_8f.sh train_tea_something_rgb_8f.sh. If you can find the trained model, it would be a great help to my research. Thanks for your reply, your help will bring hope to my desperate research. Thank you again for your help, have a nice day !!!
I was extremely desperate in my research day and night. And I have graduated this year. So I feel deep sympathy for you to do research on BAR algorithm. And I use tea50_8f as my basic model as well. It's easy for me to train the model by train_tea_something_rgb_8f.sh. I also hope to help BAR algorithm researchers like you to solve troublesome problems. Don't you worry!! Hope you can solve these issues one by one successfully.
---- Replied Message ---- | From | @.> | | Date | 06/08/2023 20:13 | | To | @.> | | Cc | @.>@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) |
Hello, I gotta check weather I save the trained model you want. If I find it I will send it to you in a few days. I am willing to help you for your research. If I can't find it, maybe I will try to train the same model on sthv1 for you. … ---- Replied Message ---- | From | @.> | | Date | 06/08/2023 17:40 | | To | @.> | | Cc | @.>@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) | Hello, would you share your trained model of TEA trained on Somethin v1 ? I need the checkpoint of TEA to pretrain my model, but it`s sooo hard for me to train TEA on Somethin v1 with not enough computing sources. It would be appreciated if you could share. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>
Oh, my God. I have never dreamed about your reply so quickly. Apologize for my late reply. Appreciate your willingness to help me with my research. Specifically, I used tea50_8f in my experiment, so I think I just need the model trained by /scrips/train_tea_something_rgb_8f.sh train_tea_something_rgb_8f.sh. If you can find the trained model, it would be a great help to my research. Thanks for your reply, your help will bring hope to my desperate research. Thank you again for your help, have a nice day !!!
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>
I was extremely desperate in my research day and night. And I have graduated this year. So I feel deep sympathy for you to do research on BAR algorithm. And I use tea50_8f as my basic model as well. It's easy for me to train the model by train_tea_something_rgb_8f.sh. I also hope to help BAR algorithm researchers like you to solve troublesome problems. Don't you worry!! Hope you can solve these issues one by one successfully. … ---- Replied Message ---- | From | @.> | | Date | 06/08/2023 20:13 | | To | @.> | | Cc | @.>@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) | Hello, I gotta check weather I save the trained model you want. If I find it I will send it to you in a few days. I am willing to help you for your research. If I can't find it, maybe I will try to train the same model on sthv1 for you. … ---- Replied Message ---- | From | @.> | | Date | 06/08/2023 17:40 | | To | @.> | | Cc | @.>@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) | Hello, would you share your trained model of TEA trained on Somethin v1 ? I need the checkpoint of TEA to pretrain my model, but it`s sooo hard for me to train TEA on Somethin v1 with not enough computing sources. It would be appreciated if you could share. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.> Oh, my God. I have never dreamed about your reply so quickly. Apologize for my late reply. Appreciate your willingness to help me with my research. Specifically, I used tea50_8f in my experiment, so I think I just need the model trained by /scrips/train_tea_something_rgb_8f.sh train_tea_something_rgb_8f.sh. If you can find the trained model, it would be a great help to my research. Thanks for your reply, your help will bring hope to my desperate research. Thank you again for your help, have a nice day !!! — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.>
Words cannot express how thankful I am. Congratulate on your graduation. I have tried a great deal of skills to improve the performance of the model on my data. Unfortunately, the effects are minimal. After a series of comparisons with the other models in MMaction2, I find that it may be a big issue without a pre-train pth of TEA. Thanks for your encouragement. I will try to continue my experiment with other skills and your shared pth file. Thank you for taking time out of your busy schedule to help me. Please accept my heartfelt thanks for your enthusiasm and good luck in your future work.
Absolutely, I got the same situation with you! Nobody can help me solve these but mysellf. I know how difficult the issues are. It's brave of you to choose BAR that needs much computing resource. Besides, I'm glad to tell you that I have found the 'ckpt.best.pth.tar' file trained by train_tea_something_rgb_8f.sh. It was trained before. And I have tested the model by single_test_tea_something_rgb_8f.sh in the TEA original code. I put the pth file in the attachment, you can download and use it directly. If something wrong with the pth file or there are any other problems, you can also email me. Hope it can help you effectively! Good luck to you too~
At 2023-06-09 15:39:54, "342419102" @.***> wrote:
I was extremely desperate in my research day and night. And I have graduated this year. So I feel deep sympathy for you to do research on BAR algorithm. And I use tea50_8f as my basic model as well. It's easy for me to train the model by train_tea_something_rgb_8f.sh. I also hope to help BAR algorithm researchers like you to solve troublesome problems. Don't you worry!! Hope you can solve these issues one by one successfully. … ---- Replied Message ---- | From | @.> | | Date | 06/08/2023 20:13 | | To | @.> | | Cc | @.>@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) | Hello, I gotta check weather I save the trained model you want. If I find it I will send it to you in a few days. I am willing to help you for your research. If I can't find it, maybe I will try to train the same model on sthv1 for you. … ---- Replied Message ---- | From | @.> | | Date | 06/08/2023 17:40 | | To | @.> | | Cc | @.>@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) | Hello, would you share your trained model of TEA trained on Somethin v1 ? I need the checkpoint of TEA to pretrain my model, but it`s sooo hard for me to train TEA on Somethin v1 with not enough computing sources. It would be appreciated if you could share. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.> Oh, my God. I have never dreamed about your reply so quickly. Apologize for my late reply. Appreciate your willingness to help me with my research. Specifically, I used tea50_8f in my experiment, so I think I just need the model trained by /scrips/train_tea_something_rgb_8f.sh train_tea_something_rgb_8f.sh. If you can find the trained model, it would be a great help to my research. Thanks for your reply, your help will bring hope to my desperate research. Thank you again for your help, have a nice day !!! — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.>
Words cannot express how thankful I am. Congratulate on your graduation. I have tried a great deal of skills to improve the performance of the model on my data. Unfortunately, the effects are minimal. After a series of comparisons with the other models in MMaction2, I find that it may be a big issue without a pre-train pth of TEA. Thanks for your encouragement. I will try to continue my experiment with other skills and your shared pth file. Thank you for taking time out of your busy schedule to help me. Please accept my heartfelt thanks for your enthusiasm and good luck in your future work.
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>
从网易163邮箱发来的超大附件推荐客户端极速下载 ckpt.best.pth.tar (187.28M, 2023年6月25日 1:25 到期) 下载
Oh, thanks for your generous help. It is lucky for me that you find the pth file. And thanks for your verification of the file, and I'm so sorry that may have taken up a lot of your time. Thanks for your pth file. Unfortunately, I couldn't find the pth download link on github or email. There are no accessible urls. Maybe I need to trouble you again to send me the download link of the pth file.
不要酱的杰哥 @.***
------------------ 原始邮件 ------------------ 发件人: "Phoenix1327/tea-action-recognition" @.>; 发送时间: 2023年6月10日(星期六) 凌晨1:27 @.>; @.**@.>; 主题: Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16)
Absolutely, I got the same situation with you! Nobody can help me solve these but mysellf. I know how difficult the issues are. It's brave of you to choose BAR that needs much computing resource. Besides, I'm glad to tell you that I have found the 'ckpt.best.pth.tar' file trained by train_tea_something_rgb_8f.sh. It was trained before. And I have tested the model by single_test_tea_something_rgb_8f.sh in the TEA original code. I put the pth file in the attachment, you can download and use it directly. If something wrong with the pth file or there are any other problems, you can also email me. Hope it can help you effectively! Good luck to you too~
At 2023-06-09 15:39:54, "342419102" @.***> wrote:
I was extremely desperate in my research day and night. And I have graduated this year. So I feel deep sympathy for you to do research on BAR algorithm. And I use tea50_8f as my basic model as well. It's easy for me to train the model by train_tea_something_rgb_8f.sh. I also hope to help BAR algorithm researchers like you to solve troublesome problems. Don't you worry!! Hope you can solve these issues one by one successfully. … ---- Replied Message ---- | From | @.> | | Date | 06/08/2023 20:13 | | To | @.> | | Cc | @.@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) | Hello, I gotta check weather I save the trained model you want. If I find it I will send it to you in a few days. I am willing to help you for your research. If I can't find it, maybe I will try to train the same model on sthv1 for you. … ---- Replied Message ---- | From | @.> | | Date | 06/08/2023 17:40 | | To | @.> | | Cc | @.@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) | Hello, would you share your trained model of TEA trained on Somethin v1 ? I need the checkpoint of TEA to pretrain my model, but it`s sooo hard for me to train TEA on Somethin v1 with not enough computing sources. It would be appreciated if you could share. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.> Oh, my God. I have never dreamed about your reply so quickly. Apologize for my late reply. Appreciate your willingness to help me with my research. Specifically, I used tea50_8f in my experiment, so I think I just need the model trained by /scrips/train_tea_something_rgb_8f.sh train_tea_something_rgb_8f.sh. If you can find the trained model, it would be a great help to my research. Thanks for your reply, your help will bring hope to my desperate research. Thank you again for your help, have a nice day !!! — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.>
Words cannot express how thankful I am. Congratulate on your graduation. I have tried a great deal of skills to improve the performance of the model on my data. Unfortunately, the effects are minimal. After a series of comparisons with the other models in MMaction2, I find that it may be a big issue without a pre-train pth of TEA. Thanks for your encouragement. I will try to continue my experiment with other skills and your shared pth file. Thank you for taking time out of your busy schedule to help me. Please accept my heartfelt thanks for your enthusiasm and good luck in your future work.
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>
从网易163邮箱发来的超大附件推荐客户端极速下载 ckpt.best.pth.tar (187.28M, 2023年6月25日 1:25 到期) 下载 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>
It's all right. I am now on my holiday and have much free time. And the pth file is upload to the attachment again. If not available still, I attach a BaiDu Netdisk link for you to download: Link:https://pan.baidu.com/s/1no-y0XEb5DB6U7xPLydVSA?pwd=2epv Password:2epv
---- Replied Message ---- | From | @.> | | Date | 06/10/2023 01:57 | | To | @.> | | Cc | @.>@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) |
Oh, thanks for your generous help. It is lucky for me that you find the pth file. And thanks for your verification of the file, and I'm so sorry that may have taken up a lot of your time. Thanks for your pth file. Unfortunately, I couldn't find the pth download link on github or email. There are no accessible urls. Maybe I need to trouble you again to send me the download link of the pth file.
不要酱的杰哥 @.***
------------------ 原始邮件 ------------------ 发件人: "Phoenix1327/tea-action-recognition" @.>; 发送时间: 2023年6月10日(星期六) 凌晨1:27 @.>; @.**@.>; 主题: Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16)
Absolutely, I got the same situation with you! Nobody can help me solve these but mysellf. I know how difficult the issues are. It's brave of you to choose BAR that needs much computing resource. Besides, I'm glad to tell you that I have found the 'ckpt.best.pth.tar' file trained by train_tea_something_rgb_8f.sh. It was trained before. And I have tested the model by single_test_tea_something_rgb_8f.sh in the TEA original code. I put the pth file in the attachment, you can download and use it directly. If something wrong with the pth file or there are any other problems, you can also email me. Hope it can help you effectively! Good luck to you too~
At 2023-06-09 15:39:54, "342419102" @.***> wrote:
I was extremely desperate in my research day and night. And I have graduated this year. So I feel deep sympathy for you to do research on BAR algorithm. And I use tea50_8f as my basic model as well. It's easy for me to train the model by train_tea_something_rgb_8f.sh. I also hope to help BAR algorithm researchers like you to solve troublesome problems. Don't you worry!! Hope you can solve these issues one by one successfully. … ---- Replied Message ---- | From | @.> | | Date | 06/08/2023 20:13 | | To | @.> | | Cc | @.@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) | Hello, I gotta check weather I save the trained model you want. If I find it I will send it to you in a few days. I am willing to help you for your research. If I can't find it, maybe I will try to train the same model on sthv1 for you. … ---- Replied Message ---- | From | @.> | | Date | 06/08/2023 17:40 | | To | @.> | | Cc | @.@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) | Hello, would you share your trained model of TEA trained on Somethin v1 ? I need the checkpoint of TEA to pretrain my model, but it`s sooo hard for me to train TEA on Somethin v1 with not enough computing sources. It would be appreciated if you could share. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.> Oh, my God. I have never dreamed about your reply so quickly. Apologize for my late reply. Appreciate your willingness to help me with my research. Specifically, I used tea50_8f in my experiment, so I think I just need the model trained by /scrips/train_tea_something_rgb_8f.sh train_tea_something_rgb_8f.sh. If you can find the trained model, it would be a great help to my research. Thanks for your reply, your help will bring hope to my desperate research. Thank you again for your help, have a nice day !!! — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.>
Words cannot express how thankful I am. Congratulate on your graduation. I have tried a great deal of skills to improve the performance of the model on my data. Unfortunately, the effects are minimal. After a series of comparisons with the other models in MMaction2, I find that it may be a big issue without a pre-train pth of TEA. Thanks for your encouragement. I will try to continue my experiment with other skills and your shared pth file. Thank you for taking time out of your busy schedule to help me. Please accept my heartfelt thanks for your enthusiasm and good luck in your future work.
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>
从网易163邮箱发来的超大附件推荐客户端极速下载 ckpt.best.pth.tar (187.28M, 2023年6月25日 1:25 到期) 下载 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>
从 网易邮箱大师 发来的云附件 |
---|
ckpt.best.pth.tar |
(187M,2023年06月25日 02:30 到期) |
下载 |
It's all right. I am now on my holiday and have much free time. And the pth file is upload to the attachment again. If not available still, I attach a BaiDu Netdisk link for you to download: Link:https://pan.baidu.com/s/1no-y0XEb5DB6U7xPLydVSA?pwd=2epv Password:2epv
---- Replied Message ---- | From | @.> | | Date | 06/10/2023 01:57 | | To | @.> | | Cc | @.>@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) |
Oh, thanks for your generous help. It is lucky for me that you find the pth file. And thanks for your verification of the file, and I'm so sorry that may have taken up a lot of your time. Thanks for your pth file. Unfortunately, I couldn't find the pth download link on github or email. There are no accessible urls. Maybe I need to trouble you again to send me the download link of the pth file.
不要酱的杰哥 @.***
------------------ 原始邮件 ------------------ 发件人: "Phoenix1327/tea-action-recognition" @.>; 发送时间: 2023年6月10日(星期六) 凌晨1:27 @.>; @.**@.>; 主题: Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16)
Absolutely, I got the same situation with you! Nobody can help me solve these but mysellf. I know how difficult the issues are. It's brave of you to choose BAR that needs much computing resource. Besides, I'm glad to tell you that I have found the 'ckpt.best.pth.tar' file trained by train_tea_something_rgb_8f.sh. It was trained before. And I have tested the model by single_test_tea_something_rgb_8f.sh in the TEA original code. I put the pth file in the attachment, you can download and use it directly. If something wrong with the pth file or there are any other problems, you can also email me. Hope it can help you effectively! Good luck to you too~
At 2023-06-09 15:39:54, "342419102" @.***> wrote:
I was extremely desperate in my research day and night. And I have graduated this year. So I feel deep sympathy for you to do research on BAR algorithm. And I use tea50_8f as my basic model as well. It's easy for me to train the model by train_tea_something_rgb_8f.sh. I also hope to help BAR algorithm researchers like you to solve troublesome problems. Don't you worry!! Hope you can solve these issues one by one successfully. … ---- Replied Message ---- | From | @.> | | Date | 06/08/2023 20:13 | | To | @.> | | Cc | @.@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) | Hello, I gotta check weather I save the trained model you want. If I find it I will send it to you in a few days. I am willing to help you for your research. If I can't find it, maybe I will try to train the same model on sthv1 for you. … ---- Replied Message ---- | From | @.> | | Date | 06/08/2023 17:40 | | To | @.> | | Cc | @.@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) | Hello, would you share your trained model of TEA trained on Somethin v1 ? I need the checkpoint of TEA to pretrain my model, but it`s sooo hard for me to train TEA on Somethin v1 with not enough computing sources. It would be appreciated if you could share. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.> Oh, my God. I have never dreamed about your reply so quickly. Apologize for my late reply. Appreciate your willingness to help me with my research. Specifically, I used tea50_8f in my experiment, so I think I just need the model trained by /scrips/train_tea_something_rgb_8f.sh train_tea_something_rgb_8f.sh. If you can find the trained model, it would be a great help to my research. Thanks for your reply, your help will bring hope to my desperate research. Thank you again for your help, have a nice day !!! — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.>
Words cannot express how thankful I am. Congratulate on your graduation. I have tried a great deal of skills to improve the performance of the model on my data. Unfortunately, the effects are minimal. After a series of comparisons with the other models in MMaction2, I find that it may be a big issue without a pre-train pth of TEA. Thanks for your encouragement. I will try to continue my experiment with other skills and your shared pth file. Thank you for taking time out of your busy schedule to help me. Please accept my heartfelt thanks for your enthusiasm and good luck in your future work.
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>
从网易163邮箱发来的超大附件推荐客户端极速下载 ckpt.best.pth.tar (187.28M, 2023年6月25日 1:25 到期) 下载 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>
从 网易邮箱大师 发来的云附件 | | | | | | | | | ckpt.best.pth.tar | | | | (187M,2023年06月25日 02:30 到期) | | | | 下载 | | | | | It's all right. I am now on my holiday and have much free time. And the pth file is upload to the attachment again. If not available still, I attach a BaiDu Netdisk link for you to download: Link:https://pan.baidu.com/s/1no-y0XEb5DB6U7xPLydVSA?pwd=2epv Password:2epv … ---- Replied Message ---- | From | @.> | | Date | 06/10/2023 01:57 | | To | @.> | | Cc | @.>@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) | Oh, thanks for your generous help. It is lucky for me that you find the pth file. And thanks for your verification of the file, and I'm so sorry that may have taken up a lot of your time. Thanks for your pth file. Unfortunately, I couldn't find the pth download link on github or email. There are no accessible urls. Maybe I need to trouble you again to send me the download link of the pth file. 不要酱的杰哥 @. ------------------ 原始邮件 ------------------ 发件人: "Phoenix1327/tea-action-recognition" @.>; 发送时间: 2023年6月10日(星期六) 凌晨1:27 @.>; @*.**@*.>; 主题: Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) Absolutely, I got the same situation with you! Nobody can help me solve these but mysellf. I know how difficult the issues are. It's brave of you to choose BAR that needs much computing resource. Besides, I'm glad to tell you that I have found the 'ckpt.best.pth.tar' file trained by train_tea_something_rgb_8f.sh. It was trained before. And I have tested the model by single_test_tea_something_rgb_8f.sh in the TEA original code. I put the pth file in the attachment, you can download and use it directly. If something wrong with the pth file or there are any other problems, you can also email me. Hope it can help you effectively! Good luck to you too~ At 2023-06-09 15:39:54, "342419102" @.> wrote: I was extremely desperate in my research day and night. And I have graduated this year. So I feel deep sympathy for you to do research on BAR algorithm. And I use tea50_8f as my basic model as well. It's easy for me to train the model by train_tea_something_rgb_8f.sh. I also hope to help BAR algorithm researchers like you to solve troublesome problems. Don't you worry!! Hope you can solve these issues one by one successfully. … ---- Replied Message ---- | From | @.> | | Date | 06/08/2023 20:13 | | To | @.> | | Cc | @.@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) | Hello, I gotta check weather I save the trained model you want. If I find it I will send it to you in a few days. I am willing to help you for your research. If I can't find it, maybe I will try to train the same model on sthv1 for you. … ---- Replied Message ---- | From | @.> | | Date | 06/08/2023 17:40 | | To | @.> | | Cc | @.@.> | | Subject | Re: [Phoenix1327/tea-action-recognition] Anyone have trained model (PTH file) of TEA trained on Kinetics 400 dataset? (Issue #16) | Hello, would you share your trained model of TEA trained on Somethin v1 ? I need the checkpoint of TEA to pretrain my model, but it`s sooo hard for me to train TEA on Somethin v1 with not enough computing sources. It would be appreciated if you could share. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.> Oh, my God. I have never dreamed about your reply so quickly. Apologize for my late reply. Appreciate your willingness to help me with my research. Specifically, I used tea50_8f in my experiment, so I think I just need the model trained by /scrips/train_tea_something_rgb_8f.sh train_tea_something_rgb_8f.sh. If you can find the trained model, it would be a great help to my research. Thanks for your reply, your help will bring hope to my desperate research. Thank you again for your help, have a nice day !!! — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.> Words cannot express how thankful I am. Congratulate on your graduation. I have tried a great deal of skills to improve the performance of the model on my data. Unfortunately, the effects are minimal. After a series of comparisons with the other models in MMaction2, I find that it may be a big issue without a pre-train pth of TEA. Thanks for your encouragement. I will try to continue my experiment with other skills and your shared pth file. Thank you for taking time out of your busy schedule to help me. Please accept my heartfelt thanks for your enthusiasm and good luck in your future work. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.> 从网易163邮箱发来的超大附件推荐客户端极速下载 ckpt.best.pth.tar (187.28M, 2023年6月25日 1:25 到期) 下载 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.> — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>
Oh, thanks for the new link bro. I have downloaded the pth file. Wish you a memorable holiday and thanks for your help!
hi @WilkyYoung @342419102 do you have a copy of something V1 dataset? I have search the Internet with no luck. Please help
Hello, I encountered low acc problem when testing the TEA model on HMDB, I only got prec@1 55.490 and prec@5 82.941, contradict with the paper that states much higher prec@1 (73.3). But, for Something v1, I got similar accuracy with the paper.
Command:
python main.py hmdb51 RGB \ --arch tea50 --num_segments 8 --gd 20 --lr 0.000125 \ --lr_steps 10 20 --epochs 25 --batch-size 8 -j 1 \ --dropout 0.8 --consensus_type=avg --eval-freq=1 \ --experiment_name=TEA3 --shift --shift_div=8 --shift_place=blockres --tune_from=./checkpoint/TEA_something_RGB_tea50_shift8_blockres_avg_segment8_e50/ckpt.best.pth.tar
Does anyone else have same issue with HMDB dataset? Thx.
EDIT: Does anyone has checkpoint of TEA trained on Kinetics 400 dataset?