The following link discusses the general rules for collecting a dataset for machine learning application:
https://www.altexsoft.com/blog/datascience/preparing-your-dataset-for-machine-learning-8-basic-techniques-that-make-your-data-better/
In addition, you can use benchmark datasets. The following links contains some good websites:
1. https://www.kaggle.com/
2. https://registry.opendata.aws/
3. https://archive.ics.uci.edu/ml/datasets.html
4. https://toolbox.google.com/datasetsearch
5. https://msropendata.com/
6. https://github.com/awesomedata/awesome-public-datasets
7. https://www.visualdata.io/
8. https://biomedicalimaging.org/2017/challenges/
Collect images from a hospital (after removing the personal information). Then ask the doctors to annotate the images (normal/abnormal, bengin/malignant, location of mass, ....).
Set of images and their information with expert (doctors) verification.
The most important question for me is
Is it enough to save image files with expert diagnose? or
save original files, expert diagnose and annotated image indicate region of interest (tumor area) and a program to calculate ROC curve for any computer algorithm perform diagnose automatic? like DRIVE data set for DR as an example
- Badges
- Science topic
- Similar topics
- Computer Science
- Images