Enhancing Biscuit Quality Control through CNN Based Defect Detection and Classification

Abstract

The problem of biscuit defect state detection can be identified as a problem in the processed food domain where the quality of the food is a crucial task. Ensuring quality of food is a crucial task and manual evaluation can be a time consuming and prune to human errors. As production processes become increasingly complex and high-speed, maintaining consistent quality become increasingly challenging. With the advancement in the technology the use of computer vision for detecting defect state may be a better approach for defect state detection which makes the process automate and less prune to human errors. Due to the production complexity current automated systems often face challenges in achieving high accuracy and adaptability across diverse production environments and conditions. A deep learning model utilizing modern CNN architecture is introduced to improve the accuracy of the modern approaches to tackle the problem. Publicly available dataset is used for the task of defect state detection. After undergoing several preprocessing techniques, the data is used to train the model for defect state detection fir optimized accuracy.