The ISIC Archive contains over 13k images of skin lesions, labeled as 'benign' or 'malignant'. The archive can be found here: https://isic-archive.com/#images

The current ways to download the database, provided by the ISIC foundation, which are known to me are the following:

1. Downloading the images one by one via the Grider API provided in the site
2. Downloading the database using their isic-archive repository in github: https://github.com/ImageMarkup/isic-archive
3. Download the database via the direct download button on their site.

The first option didn't seem feasible to us, and we had problems with the second option. The direct download in the third option never finished the downloaded successfully and always failed during the process.

We figured the direct download (which is the easiest and most comfortable way) didn't work because the file was too big and if there is a connection problem somewhere along the download process, the downloading process is stopped and isn't resumed.

So instead of downloading the entire dataset at once, you could filter the images to different sub-datasets using the filters on the site, and download each of the sub-datasets.

If you find this way too tidious, we suggest of another way: We made a script that download the entier dataset automatically. Here i'll explain the process:

Searching for a few images using the Grider api, we found that the images are stored at a url which is in the template of 'url_base_prefix + image_id + 'url_base_suffix and that their description are stored in 'url_base_prefix' + image_id

So we wrote a script that, using the Grider API, found the ids of all the images and downloaded the images and their description from the corresponding urls

You are most welcome to use the script with the simple instructions found below :)

We hope this script will allow researchers who had similliar difficulties accessing ISIC's Archive dataset and enable them to provide further work on this field, as the ISIC foundation wishes :)

1. download or clone the repository
2. Open the download_database.py file, and update the size parameter according to the updated size of the database.
3. run download.py

1. We assume that the images and descriptions have a specific url prefix and suffix as mentioned above. If the prefix or suffix ever change - let us know and we will change it accordingly! :) You can let us know in the issues tab :)

2. After first downloading the dataset we had a few images that haven't downloaded properly and were corrupted. We were not sure what was the cause of this, our guess is that there were internet connection problems during the download of these images. In any case - I would advise you to swiftly go through the images and validate that non of them are partially or entirely black (which suggests of corruption of the image) In case you experience the problem as well and find a problem with the code to be at fault please raise an issue at the issues tab and let us know :)

If you stumble into any issues - let us know in the issues section!

In addition, Any contributions of improvements to our code that will improve the comfort of the users will be dearly appreciated :)

Oren Talmor & Gal Avineri