Using pretrained when training

[huytuong010101]

According to your paper, you initialized RGB Stream weight by pretrained on ImageNet and DCT Stream by double JPEG classification. Do you think that I can train CAT-NET model without using pretrained? I reimplementing your model in tensorflow and it so difficult to load the weight of the pretrained model. Thank you <3

[CauchyComplete]

Hi again,

I'm glad to hear that someone is implementing CAT-Net in Tensorflow 😃

I am releasing results for not using pretrained weights:

The results say that:

1. Pretraining the RGB stream is almost always helpful.

2. Pretraining the DCT stream is very helpful for images that have compression artifacts but worsens the performance for images that have no compression artifacts.

3. Pretraining generally improves performance.

It's your choice whether you pretrain or not, but it's not a difficult task to do.

ImageNet is easy to obtain and the double JPEG dataset is also available publicly.

Thanks.

[huytuong010101]

Wow thank you for the helpful information <3

[dan326326]

Wow thank you for the helpful information <3 Hello, could you please tell me about the CASIA data set you used? Is there any change to the data set? I keep reporting errors and can't find the file, but the file path is correct

[huytuong010101]

Wow thank you for the helpful information <3 Hello, could you please tell me about the CASIA data set you used? Is there any change to the data set? I keep reporting errors and can't find the file, but the file path is correct

I don't use the code of this repo so I don't know what your problem I am just training with small dataset to check if model is reimplemented correctly. I use CASIAv2/Sp for training and the loss reduce to 0.24 in first 4 epoch.

[dan326326]

hello，Where did you download your CASIAv2 dataset

------------------ 原始邮件 ------------------ 发件人: "mjkwon2021/CAT-Net" @.>; 发送时间: 2021年11月11日(星期四) 中午12:29 @.>; @.@.>; 主题: Re: [mjkwon2021/CAT-Net] Using pretrained when training (Issue #4)

Wow thank you for the helpful information <3 Hello, could you please tell me about the CASIA data set you used? Is there any change to the data set? I keep reporting errors and can't find the file, but the file path is correct

I am just training with small dataset to check if model is reimplemented correctly. I use CASIAv2/Sp for training and the loss reduce to 0.24 in first 4 epoch.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.

[huytuong010101]

hello，Where did you download your CASIAv2 dataset … ------------------ 原始邮件 ------------------ 发件人: "mjkwon2021/CAT-Net" @.>; 发送时间: 2021年11月11日(星期四) 中午12:29 @.>; @.@.>; 主题: Re: [mjkwon2021/CAT-Net] Using pretrained when training (Issue #4) Wow thank you for the helpful information <3 Hello, could you please tell me about the CASIA data set you used? Is there any change to the data set? I keep reporting errors and can't find the file, but the file path is correct I am just training with small dataset to check if model is reimplemented correctly. I use CASIAv2/Sp for training and the loss reduce to 0.24 in first 4 epoch. — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.

https://www.kaggle.com/sophatvathana/casia-dataset

[dan326326]

really thank you!!!!!!!!!!!!!!!!!!!

------------------ 原始邮件 ------------------ 发件人: "mjkwon2021/CAT-Net" @.>; 发送时间: 2021年11月11日(星期四) 下午2:56 @.>; @.@.>; 主题: Re: [mjkwon2021/CAT-Net] Using pretrained when training (Issue #4)

hello，Where did you download your CASIAv2 dataset … ------------------ 原始邮件 ------------------ 发件人: "mjkwon2021/CAT-Net" @.>; 发送时间: 2021年11月11日(星期四) 中午12:29 @.>; @.@.>; 主题: Re: [mjkwon2021/CAT-Net] Using pretrained when training (Issue #4) Wow thank you for the helpful information <3 Hello, could you please tell me about the CASIA data set you used? Is there any change to the data set? I keep reporting errors and can't find the file, but the file path is correct I am just training with small dataset to check if model is reimplemented correctly. I use CASIAv2/Sp for training and the loss reduce to 0.24 in first 4 epoch. — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.

https://www.kaggle.com/sophatvathana/casia-dataset

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.