Option to train my own CNN for semantic segmentation

The existing models are trained for a specific need. As future work, we can train the CNN models to produce specific types of masks: https://missinglink.ai/guides/tensorflow/tensorflow-image-segmentation-two-quick-tutorials/

TODO:

1. Implementing flocking behaviour between the SLIC regions to determine line direction.
2. May need to implement clear edge boundaries based on flocking behaviour.
3. Change line thickness as well as line RGB values.
4. Double the size of the image.