Application research of image classification algorithm based on deep learning in household garbage sorting.

Area of Science:

• Computer Science
• Environmental Science
• Artificial Intelligence

Background:

• Manual garbage classification is error-prone and poses environmental risks.
• Automating waste classification is crucial for efficient recycling and conservation.
• Machine vision and deep learning offer potential solutions for automated waste sorting.

Purpose of the Study:

• To propose a novel deep learning-based strategy for domestic waste classification.
• To enhance the accuracy and efficiency of automated waste sorting systems.
• To reduce errors and environmental risks associated with manual waste classification.

Main Methods:

• Utilized deep learning for image-based feature extraction from waste.
• Employed the Capuchin Search Algorithm (CapSA) to optimize convolutional neural network (CNN) hyperparameters.
• Implemented a hybrid Error-Correcting Output Codes (ECOC) and Artificial Neural Networks (ANN) model for classification.

Main Results:

• Achieved high classification accuracies of 98.81% on the TrashNet dataset and 99.01% on the HGCD dataset.
• Demonstrated at least a 1.46% improvement in waste type detection compared to previous methods.
• The hybrid model's effectiveness increased with a larger number of target waste categories.

Conclusions:

• The proposed deep learning strategy is highly effective for domestic waste classification.
• The method shows significant promise for real-world applications in waste management and recycling.
• The study validates the success of the employed deep learning and hybrid model approaches.