ActMAD for Imagenet about actmad

[Yysheepsheep]

Hello, I would like to ask you about actmad, some questions I would like to ask you

[jmiemirza]

Hi @KyominHwang,

The code for ImageNet would be similar to CIFAR. You will just need to change the dataloader and you can take a pre-trained ResNet-18 from Pytorch. I would also try to clean up the code for ImageNet and push in the coming weeks, but it can take time.

@Yysheepsheep please ask. You can also write me an email if you have detailed questions. Thanks.

[KyominHwang]

Thank you for your reply! Thanks

[Yysheepsheep]

Hello, first of all thank you very much for your reply, as well as the disclosure of your code. I read your paper Dilam as well as Actmad, and the related code. I have some questions I would like to ask you. The first point is about unsupervised, the paper mentioned that you use actmad in dilam, the difference is that you use the training phase and only need a small number of unlabeled images, I want to ask is to evaluate the image for different weather conditions need to be labeled? The second point is that I see that the actmad you mentioned has no architectural requirements, so what about dilam, the detection method you used is yolo, so I can replace yolo with other detection methods to do other detection tasks.                Looking forward to your reply and thank you for your support of my research content

…

------------------ 原始邮件 ------------------ 发件人: "jmiemirza/ActMAD" ***@***.***>; 发送时间: 2023年10月19日(星期四) 下午5:32 ***@***.***>; ***@***.******@***.***>; 主题: Re: [jmiemirza/ActMAD] ActMAD for Imagenet (Issue #2) Hi @KyominHwang, The code for ImageNet would be similar to CIFAR. You will just need to change the dataloader and you can take a pre-trained ResNet-18 from Pytorch. I would also try to clean up the code for ImageNet and push in the coming weeks, but it can take time. @Yysheepsheep please ask. You can also write me an email if you have detailed questions. Thanks. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: ***@***.***>

[Yysheepsheep]

Hello, can you reply easily?

…

------------------ 原始邮件 ------------------ 发件人: "🐏🐏羊大胆🐑🐑" ***@***.***>; 发送时间: 2023年10月20日(星期五) 中午1:58 ***@***.***>; 主题: 回复： [jmiemirza/ActMAD] ActMAD for Imagenet (Issue #2) Hello, first of all thank you very much for your reply, as well as the disclosure of your code. I read your paper Dilam as well as Actmad, and the related code. I have some questions I would like to ask you. The first point is about unsupervised, the paper mentioned that you use actmad in dilam, the difference is that you use the training phase and only need a small number of unlabeled images, I want to ask is to evaluate the image for different weather conditions need to be labeled? The second point is that I see that the actmad you mentioned has no architectural requirements, so what about dilam, the detection method you used is yolo, so I can replace yolo with other detection methods to do other detection tasks.                Looking forward to your reply and thank you for your support of my research content

------------------ 原始邮件 ------------------ 发件人: "jmiemirza/ActMAD" ***@***.***>; 发送时间: 2023年10月19日(星期四) 下午5:32 ***@***.***>; ***@***.******@***.***>; 主题: Re: [jmiemirza/ActMAD] ActMAD for Imagenet (Issue #2) Hi @KyominHwang, The code for ImageNet would be similar to CIFAR. You will just need to change the dataloader and you can take a pre-trained ResNet-18 from Pytorch. I would also try to clean up the code for ImageNet and push in the coming weeks, but it can take time. @Yysheepsheep please ask. You can also write me an email if you have detailed questions. Thanks. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: ***@***.***>