LCD Panel Defect Detection using Unsupervised learning

A total of 3 people were involved in this project. The main challenge of this project was to detect the defect on the LCD panel where the number of training samples in the dataset was very few(couple hundred samples). My role was to analyze the dataset and write the necessary python codebase to train the deep learning models and the inference. In totality, we created an application that checks all the LCD panel images in a folder and shows the defects of the LCD panel.

This project was given to us by the client as part of an AI plus competition and though we were a small team, this motivated us to go and finish the project in just two months.

Timeline: 2 months. Awards: Secured second place in the AI plus competition. Deliverables: Achieved • 0% Overkill • 96% Accuracy