Bag of Features

Python Implementation of Bag of Words for Image Recongnition using OpenCV and sklearn

Original author

This is a clean fork of bikz05's work.

Installation

Dependencies:

• Python (>= 3.3)
• Numpy (>= 1.6.1)
• Scipy (>= 0.9)
• Scikit Learn

Note

I am using OpenCV3 and a number of things have been moved to the opencv_contrib repo. Make sure you install the xfeatures2d module to be able to use SIFT.

Training the Classifier

python learn.py -t dataset/train -c dataset/model

Testing the Classifier

• Testing a single image

python predict.py -t dataset/test/class/image.extension --show

• Testing a testing dataset

python predict.py -t ../dataset/test/ -m ../dataset/model -r ../dataset/results --verbose

The --visualize flag will display the image with the corresponding label printed on the image/

Troubleshooting

If you get

AttributeError: 'LinearSVC' object has no attribute 'classes_'

error, the simply retrain the model.