The hand imagery database: a comprehensive collection of images (both clean hand shots and background shots) for the training of Haar Cascade Classifiers.
This site hosts a growing database of one thing, and one thing only: images of hands. The reason for this is purely, for science! Or more specifically, to help develop motion-tracking software which uses object recognition.
The database is needed to train object-detection algorithms to detect hands with little margin for error. To train the Haar algorithms well you need a database of literally thousands of images. So if you just want to lend us a hand, we suggest you get to it straight away!
If you want to read more about the reason behind this database, have a read of this page.
The database is for open-source purposes. It will be fully available to anyone who wants it. So if you don't want to show your face (or your hands), don't submit your images to this site.
That being said, the database is completely anonymous - no names attached to images. We hope you can contribute! (:
- Total number of images:
- Total filesize:
- Number of positive images:
- Number of background images:
- Number of peer-submitted classifiers:
A longer explanation as to what it could be used for is on the About page.
If you'd like to see what I myself am doing with it, you can take a look here.