search

huyquoctrinh / MetaPolyp-CBMS2023

Meta-Polyp: a baseline for efficient Polyp segmentation (CBMS 2023)
MIT License
68 stars 9 forks source link

doubt about datasets setting #2

Closed ShawnHuang497 closed 11 months ago

ShawnHuang497 commented 1 year ago

Hi! U said that "with training set from 50% of Kvasir-SEG and 50% of ClinicDB dataset" in the README.md, but U said "The training, validating, and testing make up 60%, 20%, and 20%, respectively".

So I'm confused about your datasets division setup, can you answer that?

huyquoctrinh commented 1 year ago

Thank you so much for this issue, data was gotten from the PraNet, and the training set is half of Kvasir-SEG and ClinicDB, then for the training stage, I split it 60:20:20 for train:val:test, and then I will use the test dataset from PraNet for benchmark the model. I hope this answer satisfy your question.

ShawnHuang497 commented 12 months ago

Ok, thanks for your answer. And I am planing to follow your work, but the dataset link of Pranet is invalid. Can you send the splited datasets .zip or name list to me? My email is 2515950785@qq.com. Thanks.

huyquoctrinh commented 12 months ago

Hi, I have updated the link to download dataset on the README, I hope you can find that

ShawnHuang497 commented 12 months ago

Great job! Thanks very much! Moreover, I do not see the test results of ClinicDB in your paper. Will you be releasing the results of ClinicDB test dataset?

huyquoctrinh commented 11 months ago

Hi, because at that time I don't have enough pages for writing, I have to choose 3 datasets for benchmark results...