AIVResearch / MSANet

Official Pytorch implementation of Multi-Similarity and Attention Guidence for Boosting Few-Shot Segmentation.
https://arxiv.org/abs/2206.09667v1
63 stars 12 forks source link

coco dataset #9

Open Galaxy-PC opened 1 year ago

Galaxy-PC commented 1 year ago

Hello, I want to ask the data processing part: how to preprocess the coco dataset, how to set the file directory, in addition, SharePoint I can't log in, can't get base_annotations (what is this file?) content. Thank you! QQ截图20230423171822 QQ截图20230423171852

Ehteshamciitwah commented 1 year ago

Hello, Sorry for the confusion.

Base annotations are the files used for the base learner. We used the same base learner annotation as in BAM. The data preprocessing is also the same as BAM. I update the path of the base annotation. You can download it directly from the BAM repository.

Galaxy-PC commented 1 year ago

Thanks for the reply.

As an undergraduate, I am learning from your excellent results. According to your prompt I downloaded the base_annotations from BAM, this should be a label for the coco dataset, but I still get an error running it. I found that he reads the val file but wants to read the image of the train, and the reading is wrong, whether this is where the code needs to be modified, or how the dataset file is prepared. 

Thank you!

希奇儿 @.***

 

------------------ 原始邮件 ------------------ 发件人: "AIVResearch/MSANet" @.>; 发送时间: 2023年4月24日(星期一) 上午9:04 @.>; @.**@.>; 主题: Re: [AIVResearch/MSANet] coco dataset (Issue #9)

Hello, Sorry for the confusion.

Base annotations are the files used for the base learner. We used the same base learner annotation as in BAM. The data preprocessing is also the same as BAM. I update the path of the base annotation. You can download it directly from the BAM repository.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

Galaxy-PC commented 1 year ago

Hello, I would like to know what the dataset file path setting looks like? The code is how data is preprocessed, and I run it in Windows 10 and can't see exactly what went wrong.

Galaxy-PC commented 1 year ago

Hello, I found several problems: 1.lists\coco\fss_list\val\ the txt file inside is a train-related dataset, and there will be problems with reading. 2. Load the model file weights, there is a mismatch between the key-value pairs, whether there is a problem with the file. 3.The MSHNet/config of the Scripts in the readme is not misspelled. 屏幕截图 2023-04-25 001857 屏幕截图 2023-04-25 001953