sky-detector

A non-deep learning sky detector method with simple python script

Intro

This project use very simple algorithm, we assume that the variance of sky is smaller than non-sky part. Therefore, we use Laplacian filter to extract the part that variance is lower than threshold.

After we extract the mask after filtering out non-sky part, then we examine every single skyline, and find out the starting point that is non-sky part.

This method was inspired by the papaer Sky Region Detection in a Single Image for Autonomous Ground Robot Navigation

Requirements

opencv-python = "^4.2.0"
scipy = "^1.5.0"
numpy = "^1.19.0"

Run the test

poetry shell # We use poetry to run the whole script, install dependencies
python test.py

Demo