Related Experiment Video
Updated: Sep 7, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
632
Deep convolutional neural networks for construction and demolition waste classification: VGGNet structures, cyclical
Kunsen Lin1, Tao Zhou2, Xiaofeng Gao3
1The State Key Laboratory of Pollution Control and Resource Reuse, College of Environmental Science and Engineering, Tongji University, 1239 Siping Road, Shanghai, 200092, China.
Journal of Environmental Management
|June 19, 2022
Summary
This study introduces CVGGNet models for automated Construction and Demolition (C&D) waste sorting using deep learning. CVGGNet-16 achieved the best performance, improving waste management and circular economy initiatives.
Area of Science:
- Computer Science
- Environmental Science
- Engineering
Background:
- Construction and Demolition (C&D) waste sorting is crucial for circular economy development.
- Automated sorting minimizes human contact, reducing pathogen spread and COVID-19 contamination risks.
Purpose of the Study:
- To develop an efficient deep learning method for sorting C&D waste images.
- To utilize knowledge transfer and data augmentation for improved classification accuracy.
Main Methods:
- Proposed four CVGGNet models (VGGNet-11, -13, -16, -19) based on knowledge transfer.
- Employed data augmentation and cyclical learning rates for training.
- Utilized t-distributed Stochastic Neighbor Embedding (t-SNE) for dimensionality reduction and visualization.
Main Results:
- CVGGNet-16 achieved the highest accuracy (76.6%), weighted average precision (76.8%), recall (76.6%), F1-score (76.6%), and micro ROC (87.0%).
- Knowledge transfer reduced training time for CVGGNet-11 from 1039.45s to 991.05s and improved its performance.
- Training time increased with model depth, with CVGGNet-19 taking the longest (1337.81s).
Conclusions:
- CVGGNet models demonstrate strong performance in automatically sorting C&D waste.
- The developed method offers a pathway towards enhanced C&D waste management.
- t-SNE effectively visualizes distinct separation of C&D waste types.
