HawkEye

Mobile application for analysis of solar energy potential in an urban setting for the Indian Space Research Organization (ISRO)

Winner, college qualifier round, Smart India Hackathon 2020

Problem Statement

• To develop a mobile application for detecting sky pixels in a photograph
• Generate mask image (sky pixels - white, other pixels - black)
• Calculate angle of elevation of lowest sky pixel in mask image

Features

• Machine Learning Model - HRNet - for semantic segmentation of landscape images
• Achieved a misclassification rate of 4.7% on a sample of 500 images from the skyfinder dataset
• Find angle of elevation of lowest sky pixel with Canny edge detection, HOG transform, thresholding and trigonometry using details from the camera's optics (proposal)