1

I want to test some clustering algorithms on a handwritten digit dataset. Is there any handwritten digit dataset with an already extracted features ? I'm not searching for a handwritten digit dataset of brute image pixels like the case of MNIST for example. I'm searching for ready vectors of features that are already extracted from such a dataset.

João Almeida
  • 4,487
  • 2
  • 19
  • 35
shn
  • 5,116
  • 9
  • 34
  • 62
  • What do you mean by "ready vectors of features" ? Do you mean, computer vision "features" like SIFT, Harr etc ? – sumodds Feb 09 '12 at 20:52
  • Sorry, I am not aware of any. The most famous ones in handwritten character recognition is the Yann Lecun's MNIST and USPS data set (available on Sam Roweis's site). But why don't you just run SIFT or a simple run length on it, it won't take too long. Doing with opencv / MATLAB isn't too time consuming either. – sumodds Feb 11 '12 at 18:58

2 Answers2

3

Have a look at the UCI Machine Learning Repository. IIRC it has at least two handwritten digit datasets.

Has QUIT--Anony-Mousse
  • 76,138
  • 12
  • 138
  • 194
0

Here you have an example of a handwritting dataset (and another one) of the link @Anony-Mousse gave you (I think that's the first site you had to go).

And here's another example I found. Just use Google, he's your best friend in most cases :)

Jose Garrido
  • 732
  • 1
  • 15
  • 31