We will detect changes in images like cracks, color fading or missing parts. it can compare two or more images and highlight what changed. we used opencv and numpy for image subtraction and contour detection. then we added a simple cnn model (mobilenet or resnet) to classify the type of change. the results can be shown in a web dashboard made with streamlit or flask.
the idea came from real world inspection problems like checking tire wear or color fading on signs. humans cant see small slow changes properly so we wanted to make a system that can do it automatic and more accurate.
it will help us learn image processing, ai models and how to combine them to get better results.
Challenges
images taken in different light or angles gave wrong results.
hard to make model understand what is real crack vs just shadow.
took time to clean and resize images for cnn input.
Log in or sign up for Devpost to join the conversation.