Real-Time Detection of Waste Battery Defect using Synthetic Image Datasets based on Deep Learning

The growing electric vehicle market worldwide is leading to an increase in waste batteries, posing significant environmental and safety challenges. This paper introduces an innovative framework for real-time defect detection, aimed at enhancing the efficiency and accuracy of waste battery sorting, thereby overcoming the drawbacks of traditional manual inspection methods. The proposed framework encompasses three pivotal components: the generation of 3D models from real-world data, the creation of synthetic datasets via a game engine, and the deployment of deep learning algorithm for automated inspection. This integrated approach optimizes the battery sorting process, substantially reducing time and labor costs. The most advanced part of the framework is a deep learning model that uses simulated synthetic data to quickly identify defects in waste batteries in real-time. The results underscore the framework