Shadow Detector

Installing Dependencies

pip install -r requirements.txt
git submodule update --init
cd pymeanshift
./setup.py install

Downloading SBU-Dataset and setting up project structure

./init.sh

Running this will download the SBU shadow dataset and extract it under the data directory, as well as create the checkpoints directory.

Training the model

./train.py

This will save the models with the best accuracies under the checkpoints directory.

Detecting shadows

To pass the whole image to the CNN and find the segments which are shadows:

./detect_shadows.py --image <path to image> --model <path to model>

To pass individual segments to the CNN and find the which are shadows:

./detect_shadows_by_segment.py --image <path to image> --model <path to model>

<path to image> is the path to any image.

<path to model> is the to either model.h5 or any of the models under the checkpoints directory.