I split my dataset into 10 classes, and I set init cls to 10, again reporting IndexError: Target 853 is out of bounds. How should this kind of mistake be solved

Question

I split my dataset into 10 classes, and I set init cls to 10, again reporting IndexError: Target 853 is out of bounds. How should this kind of mistake be solved

BLACKJACK0711 opened this issue a year ago · 5 comments

Answer 1 · 2023-06-17T02:05:40.000Z

Hi, could you please share your folder architecture of your dataset?

Answer 2 · 2023-06-19T05:47:20.000Z

Hi, could you please share your folder architecture of your dataset?

Thank you for your concern about this issue. Please check your email for relevant introduction

Answer 3 · 2023-06-20T03:04:58.000Z

Hi, thanks for your introduction! The datasets used in this repo are image datasets and the data folders of them are organized like following.

dataset main folder
├── train
    ├── class 0
        ├── class 0 train image 0
        ├── class 0 train image 1
        ├── ...
    ├── class 1
        ├── class 1 train image 0
        ├── class 1 train image 1
        ├── ...
    ├── ...
├── test
    ├── class 0
        ├── class 0 test image 0
        ├── ...
    ├── class 1
        ├── class 1 test image 0
        ├── ...
    ├── ...

With such folder architecture, the data files will be correctly read by torchvision.datasets.ImageFolder. You can find details from dataset loading classes called iImageNet1000, iImageNet100 and iVGGFace200 in utils/data.py.

To run this repo, please make sure your dataset samples are images and they are put in correct folders.

Answer 4 · 2023-06-20T15:14:35.000Z

I converted my data into images through code, I am not sure what is the reason for the error，I added the print function below the logits line to see what its output looks like： Its result is this： Finally, please take a look at my specific error：  

…

------------------ 原始邮件 ------------------ 发件人: "xfflzl/CIM-CIL" ***@***.***>; 发送时间: 2023年6月20日(星期二) 中午11:05 ***@***.***>; 抄送: "BLACK ***@***.******@***.***>; 主题: Re: [xfflzl/CIM-CIL] I split my dataset into 10 classes, and I set init cls to 10, again reporting IndexError: Target 853 is out of bounds. How should this kind of mistake be solved (Issue #6) Hi, thanks for your introduction! The datasets used in this repo are image datasets and the data folders of them are organized like following. dataset main folder ├── train ├── class 0 ├── class 0 train image 0 ├── class 0 train image 1 ├── ... ├── class 1 ├── class 1 train image 0 ├── class 1 train image 1 ├── ... ├── ... ├── test ├── class 0 ├── class 0 test image 0 ├── ... ├── class 1 ├── class 1 test image 0 ├── ... ├── ... With such folder architecture, the data files will be correctly read by torchvision.datasets.ImageFolder. You can find details from dataset loading classes called iImageNet1000, iImageNet100 and iVGGFace200 in utils/data.py. To run this repo, please make sure your dataset samples are images and they are put in correct folders. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

Answer 5 · 2023-06-21T07:15:38.000Z

Hi, it seems that the images are removed in the email.