5th World Conference on Smart Trends in Systems, Security and Sustainability

Authors - Harshith MohanKumar, Mohammed Zayd Jamadar, S Natarajan, Nithin H
Abstract - Detecting and classifying recyclable material is one of the hardest computer vision problems. In this study we present an optimized residual network with the inclusion of an SVM classification layer to categorize single object images containing recyclable waste. The model is trained on three thousand images of which 1/3rd were of recyclable products commonly found in India. The results of the trained model have shown to contain competitive accuracy rates and reliably fast inference times.