Downloading data for the ISIC experiments

laura-rieger / deep-explanation-penalization

Code for using CDEP from the paper "Interpretations are useful: penalizing explanations to align neural networks with prior knowledge" https://arxiv.org/abs/1909.13584

MIT License

127 stars 14 forks source link

Downloading data for the ISIC experiments #3

Closed expectopatronum closed 4 years ago

expectopatronum commented 4 years ago

Hi! I am trying to download the data for the reported ISIC experiments using the scripts you provided. I encountered the following issues:

00_download_metadata.py prints Fetching metadata for 23906 images. The paper states The ISIC dataset consists of 21,654 images (19,372 benign). How can I figure out which images you used?
01_download_imgs.py only downloads 50 images although in the config the value is set to 25000. Am I expected to change start_offset = 0 or the limit parameter in the URL somehow?

Thanks a lot for the great work! Best regards Verena

laura-rieger commented 4 years ago

Hi Verena,

glad that you liked the work! We excluded images with an indeterminate or unknown diagnosis so only images that were diagnosed as either 'benign' or 'malignant' were included (see 02_sort_imgs.py l.38-44). We'll make this more clear in the final paper, thanks for pointing it out.

The start_offset is counted up in l.31 in script 01_download_imgs.py, so it should work as it is. Did the script finish after 50 images or was it hanging? I'll look into it in any case.

Best, Laura

laura-rieger commented 4 years ago

It should be fixed now, there was a weird typo in l. 18 (used modulo instead of division).

expectopatronum commented 4 years ago

Thanks! It works! I am not sure if I should/can close it since you reopened it :)

laura-rieger commented 4 years ago

Super!

Juna2 commented 4 years ago

Hi, again. I'm trying to download the ISIC dataset with 01_download_imgs.py file but it takes too much time to download it.(more than 9 hours sometimes 11 hours) Is there something that I missed? And at the ISIC site, they have more than 25000 images. so I don't know which images to download. Any advice?

laura-rieger commented 4 years ago

Hi again, I'm fairly sure the downloading speed is an issue on their side and I don't know of a way to resolve it. I followed their recommendations for the API, so they may have updated it in the meantime.

If you go to the gallery you'll find that some images don't have a diagnosis or don't have a definite diagnosis. The script won't include those images in the dataset so you don't need to choose which images to download.

Juna2 commented 4 years ago

I eventually managed to download them all haha In order to make it quick, I terminated the downloading process and start again a few times. And when I start again I skipped all previously downloaded files. The process tends to slow down over time.