Handwritten digit recognizer implemented with HOG + SVM.

I recommend you have virtualenv installed. After that, create a new environment and activate it:

$> virtualenv -p python3 venv
$> source venv/bin/activate

After that, just install the requirements:

pip install -r requirements.txt

At the root of the project, execute the following command to train a model:

python train.py -d ./data/digits.csv -m ./models/svm.cpickle

And then run your model on some image, like this:

python classify.py -i images/umbc_zipcode.png -m models/svm.cpickle

You'll get an output in the terminal similar to this:

I think that number is: 2
I think that number is: 1
I think that number is: 2
I think that number is: 5
I think that number is: 0

And an annotated photo, like this: