Project Details

Used Laplace of the Variance as the sole feature to train MLP (MultiLayer Perceptron) that can determine if an image is blurred or not, with an accuracy of 83.04%.

Dataset used: CERTH image blur dataset found here

Steps to install and run

1. Download and install the latest version of Anaconda (Python 3.6) from here(https://www.anaconda.com/download) (Will include numpy, pandas and scikit-learn by default).
2. Both pip and conda are included in Anaconda and Miniconda, so you do not need to install them separately.
3. Install latest version of OpenCV using pip install opencv-python
4. Run script (both train and test script included) using python blur_detector.py and wait for the prompt to stop.