Improving the performance of CNN by transfer learning for waste classification
Israa Nasir Abood, Ghaidaa Abdul Aziz Al-Talib · AIP conference proceedings · 2025
The output of home garbage has rapidly increased as a result of fast economic and social growth.Intelligent waste classification techniques are becoming a crucial part of how people can attain sustainable development.The accuracy and efficiency of conventional waste classification techniques are not efficient.To increase the processing of waste classification, deep learning approaches have been introduced in this field as they deal with images rather than features.Convolutional Neural Networks (CNNs) are one of the most popular deep learning architectures in the field of computer vision.However, training CNNs from scratch can be computationally expensive and requires a large amount of labeled data.Transfer learning is a technique that addresses these challenges by leveraging pre-trained models on large datasets to improve the performance of other models using only smaller datasets.This paper employed deep learning and transfer learning for waste classification on a specific data set of waste images, as well as proved that the transfer learning approach can be an effective method for waste classification.This paper proposed an improved model to reduce human intervention and rely more on technology for a more productive and cost-effective process.Firstly, the CNN was trained with a dataset that is available on the Kaggle website and obtained an accuracy of about 88% for classifying six classes of waste.Secondly, the learning of CNN was transferred to evaluate and test four algorithms which are VGG16, InceptionV3, MobileNetV2, and EfficientNetB0 that achieved accuracies of 92%, 94 %, 95%, and 97% respectively.The results showed that transfer learning significantly improves the accuracy and reduces the training time of CNNs for waste classification.