This paper investigates the methods and results of building a computer vision-based object recognition system in waste recycling processes. The study evaluated the performance of a CNN model using a dataset of 19,762 images with 10 classes, including clothing, metal, glass, biological waste, and other recyclable materials. The model was trained for 30 epochs and stopped at the 23rd epoch using the Early Stopping mechanism, achieving a training accuracy of 95.57% and a validation accuracy of 81.68%. The results showed that the model had high training efficiency, but the low validation accuracy limited its generalization ability. The data distribution imbalance and overtraining symptoms indicated the need for additional augmentation and optimization. The study confirmed the potential of computer vision in automating waste sorting, but the model needs further development for real-time application. Future research is recommended to expand the dataset, use hybrid models, and optimize in real time. The results lay the foundation for the development of innovative solutions that contribute to environmental protection and efficient use of resources.
Aben et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: