ML/Deep Learning - Where do you get your datasets?
Where do you get datasets for training. I mean specific things like Lung CTs with diseases or research datasets
Where do you get datasets for training. I mean specific things like Lung CTs with diseases or research datasets
Google Dataset Search is useful - https://datasetsearch.research.google.com/
99% of the challenge of training deep learning image classifiers is data acquisition, so if you find a good shortcut, you most likely have a great product on your hands.
In my case, we paid a team of doctors and a team of volunteer patients to produce the required data. I believe it came out at roughly €1000 per person scanned.
But for lung scans specifically, I believe the university of Geneva has a dataset funded from EU budgets that they used for developing their radiology image retrieval application as part of the KnowARC project. If you're located inside the EU, they should grant you access.
For a final project in an ML class i just contacted several medical departments via E-Mail and asked for advice how to acquire the specific data i was looking for. Luckily one of the departments sent me a spreadsheet containing exactly the data i was looking for.
Maybe try your luck with that.
UCI ML Repo
I have requested research labs previously for very niche datasets and usually they will give you. You just have to find which dataset you need.
But they are super big as well, think terabytes of image data.
This comment was deleted 6 years ago