Trained and evaluated a stacked CNN architecture for rooftop image classification
Extract the Images.zip file to get the images for various rooftops.
This is a classification problem with 4 types of roofs (Gable, Hip, Mix, and Flat). Only the probabilities over 0.65 are finalized, the predictions with probabilities less than 0.65 are asked for manual classification. The manual inputs are later used for future model improvement.