search

lofrienger / S2ME

S2ME: Spatial-Spectral Mutual Teaching and Ensemble Learning for Scribble-supervised Polyp Segmentation (MICCAI 2023)
15 stars 1 forks source link

Hello,in the data preparing file, I find there si a 'csv_root',what data should i put into it? #6

Open yike5022 opened 4 months ago

lofrienger commented 4 months ago

image It is the data split file in data/polyp. Sorry for the naming confusion that it is in json but not csv format.

yike5022 commented 4 months ago

could you please tell we how to get this json file?because i really want to see the testing picture fo the SUN-SEG dataset,thank you

yike5022 commented 4 months ago

oh i just see the file,thanks!

yike5022 commented 4 months ago

could you please offer the SUN-SEG testing pictures? so I can see the pictures directly ,my code ability is so poor

lofrienger commented 4 months ago

Sorry, but I can't help. I don't have the right to share the dataset with you. You should follow the instructions from the dataset owner to get access to the test data. FYI: https://github.com/GewelsJI/VPS/blob/main/docs/DATA_PREPARATION.md

yike5022 commented 4 months ago

thanks! i have already get the access!may i ask you how to save the testing binary maps? because i have no idea to get the testing pitcture path to save them.

lofrienger commented 4 months ago

Hi, after argmax at the output layer, you should get the binary map. Then you can save it (batch size=1) or them (batch size>1) with torch or opencv (cv2.imwrite()). I am not sure what you mean by "the testing pitcture path to save them". Basically, you can save to wherever you like.

yike5022 commented 4 months ago

oh thank you! i know what you mean!Thanks for your reply again!

yike5022 commented 4 months ago

Hello, I still have a question, how to save the test data in the format of the original data set, because then I can use the evaluation indicators provided by the VPS to test the indicators of your code

lofrienger commented 4 months ago

Perhaps you should convert the shape of the saved masks. Please check which type of masks you need, like RGB or grayscale. Make sure the saved masks and the ground truth masks have the same channel number (1 or 3).