Closed hyoje42 closed 4 years ago
Hi,
Thanks
When I use ResNet18 as the backbone with a 80x80 input, it produce a error:
it seems the code is not correct, so how to reproduce your result when taking as input a 80x80 image but 224x224? @nupurkmr9
When I use ResNet18 as the backbone with a 80x80 input, it produce a error:
it seems the code is not correct, so how to reproduce your result when taking as input a 80x80 image but 224x224? @nupurkmr9
Have you solved this problem?
Dooray! Failure Notice
Failure Notice
Your message sent to
@.*** has failed to be delivered.
Please refer to the below for details.
* Recipient :
@.***
* Sent time :
2024-03-28T23:26:54
* Subject :
Re: [nupurkmr9/S2M2_fewshot] Did you use 80*80 images for mini-Imagenet dataset? (#4)
* Remote host said :
대외비/기밀 메일은 외부 계정으로 발송이 제한되어 일부 메일 발송 실패(수신자가 전달 설정한 경우 포함)
This message was sent from a notification-only address that cannot accept incoming email.
For more information, please contact ***@***.***
© Dooray!.
Hi. Thanks for sharing your code. Your paper and code help a lot for me. I have some questions.
Did you use 8080 images for mini-Imagenet dataset? I read your code, so I thought you used 8080 cropped image for mini-Imagenet, but I want to clarify it. From the issue at 'A Closer Look At Fewshot', Chen, https://github.com/wyharveychen/CloserLookFewShot/issues/20 I think that Chen used 224224 images for experiments that use Resnet as backbone. (I thought it is unfair to other papers, because original mini-Imagenet dataset consists of 8484 images. If I were wrong, please correct me)
Where can you get the results on CIFAR dataset over Baseline++ with WRN-28-10 as backbone? Chen did not use WRN-28-10 and CIFAR dataset, as far as I know. Did you get the results by yourself? Also I found that the reported accuracy little bit different from Chen's results. For the mini-Imagenet dataset(5shot-5way), 76.16+-0.63 from Chen, but 74.14+-0.19 from yours. In your paper, did you get baseline++'s results by yourself?
Thanks, in advance.