This document presents a computer vision-based system for early fire detection using machine learning. It captures video using a webcam and processes the frames to detect fire. The frames are converted to different color models and compared using OpenCV. Morphological transformations and thresholding are applied to detect fire regions. If fire is detected, an alarm is triggered and a notification email is sent. The system aims to reduce limitations of conventional fire detection methods and optimize detection using image processing techniques. A literature review is also presented discussing previous works on fire detection using computer vision and machine learning.