acon icon indicating copy to clipboard operation
acon copied to clipboard

Experimental results on CIFAR-100

Open jiequancui opened this issue 3 years ago • 9 comments

Hi, Thanks for your nice work.

From the paper, the improvements on ImageNet are obvious, which is attractive to me. Recently, I conduct experiments on CIFAR-100 with ResNet-18 and the meta-ACON. However, the performance is not satisfactory. Here, I want to know if you can provide some results on CIFAR-100 for reference.

Very thanks.

jiequancui avatar Aug 27 '21 12:08 jiequancui

hi ,Have you solved the problem yet?can you give me some advice?

cheun726 avatar Sep 03 '21 03:09 cheun726

Unfortunately, I haven't solved the problem yet. I'm still trying it on ImageNet. I tried on CIFAR-100 with ResNet-18 and meta-ACON. The performance is still far behind the original ResNet-18. I am wondering if the configuration of my meta-ACON-ResNet18 is correct. I replace all the relu with meta-acon in all the basic block.

jiequancui avatar Sep 03 '21 04:09 jiequancui

Have you ever run a program provided by the author?

I'm using IMAGENET dataset for the first time。

I don't know what this script means。https://raw.githubusercontent.com/soumith/imagenetloader.torch/master/valprep.sh

can you give me some advice?

------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2021年9月3日(星期五) 中午12:00 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [nmaac/acon] Experimental results on CIFAR-100 (#13)

Unfortunately, I haven't solved the problem yet. I'm still trying it on ImageNet. I tried on CIFAR-100 with ResNet-18 and meta-ACON. The performance is still far behind the original ResNet-18. I am wondering if the configuration of my meta-ACON-ResNet18 is correct. I replace all the relu with meta-acon in all the basic block.

— 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.

cheun726 avatar Sep 03 '21 04:09 cheun726

Hi, As seen from the script, it just put images in the validation set into correct folders. The folder named with code, like n01751748, just represents a specific class.

jiequancui avatar Sep 03 '21 04:09 jiequancui

thank you. How do I apply meta-ACON to the network structure of other tasks?for example super-resolution。Could you give me some advice?

------------------ 原始邮件 ------------------ 发件人: "nmaac/acon" @.>; 发送时间: 2021年9月3日(星期五) 中午12:27 @.>; @.@.>; 主题: Re: [nmaac/acon] Experimental results on CIFAR-100 (#13)

Hi, As seen from the script, it just put images in the validation set into correct folders. The folder named with code, like n01751748, just represents a specific class.

— 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.

cheun726 avatar Sep 03 '21 04:09 cheun726

Hi, I'm no familiar with the super-resolution task.

jiequancui avatar Sep 03 '21 04:09 jiequancui

thank you

------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2021年9月3日(星期五) 中午12:50 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [nmaac/acon] Experimental results on CIFAR-100 (#13)

Hi, I'm no familiar with the super-resolution task.

— 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.

cheun726 avatar Sep 03 '21 04:09 cheun726

hi .

How do I set the value of --val-dir?Why did this error occur?can you give me some advice ?

------------------ 原始邮件 ------------------ 发件人: "nmaac/acon" @.>; 发送时间: 2021年9月3日(星期五) 中午12:50 @.>; @.@.>; 主题: Re: [nmaac/acon] Experimental results on CIFAR-100 (#13)

Hi, I'm no familiar with the super-resolution task.

— 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.

cheun726 avatar Sep 04 '21 04:09 cheun726

hello,

Do you know how to replace relu in network with metaacon?can you give me some advice?

------------------ 原始邮件 ------------------ 发件人: "nmaac/acon" @.>; 发送时间: 2021年9月3日(星期五) 中午12:50 @.>; @.@.>; 主题: Re: [nmaac/acon] Experimental results on CIFAR-100 (#13)

Hi, I'm no familiar with the super-resolution task.

— 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.

cheun726 avatar Sep 12 '21 07:09 cheun726