A Study on the Effectiveness of CNN-Based Summarization in Small-Target Waste Sorting: Comparative Analysis of EfficientNet, ViT, and ConvNeXt

Zeyu Shen

doi:10.47297/taposatWSP2633-456930.20260702

PDF
导出
分享
收藏
专辑

A Study on the Effectiveness of CNN-Based Summarization in Small-Target Waste Sorting: Comparative Analysis of EfficientNet, ViT, and ConvNeXt

7卷第2期

浏览量: 0 Downloads: 0

A Study on the Effectiveness of CNN-Based Summarization in Small-Target Waste Sorting: Comparative Analysis of EfficientNet, ViT, and ConvNeXt
A Study on the Effectiveness of CNN-Based Summarization in Small-Target Waste Sorting: Comparative Analysis of EfficientNet, ViT, and ConvNeXt
School of Computing and Data Science, Xiamen University Malaysia, Xiamen, Fujian, 43900, China
2026年7卷第2期页码：151-154
纸质出版：2026-02-28
DOI：10.47297/taposatWSP2633-456930.20260702
Accepted：

Scan QR Code

Zeyu Shen. (2026). A Study on the Effectiveness of CNN-Based Summarization in Small-Target Waste Sorting: Comparative Analysis of EfficientNet, ViT, and ConvNeXt. 科技理论与实践（英文版）, 7(2), 151-154.

Shen Zeyu.A Study on the Effectiveness of CNN-Based Summarization in Small-Target Waste Sorting: Comparative Analysis of EfficientNet, ViT, and ConvNeXt[J].Theory and Practice of Science and Technology,2026,07(02):151-154.
Zeyu Shen. (2026). A Study on the Effectiveness of CNN-Based Summarization in Small-Target Waste Sorting: Comparative Analysis of EfficientNet, ViT, and ConvNeXt. 科技理论与实践（英文版）, 7(2), 151-154. DOI： 10.47297/taposatWSP2633-456930.20260702.

Shen Zeyu.A Study on the Effectiveness of CNN-Based Summarization in Small-Target Waste Sorting: Comparative Analysis of EfficientNet, ViT, and ConvNeXt[J].Theory and Practice of Science and Technology,2026,07(02):151-154. DOI： 10.47297/taposatWSP2633-456930.20260702.

摘要

Abstract

Precise classification of small objects poses a critical challenge for computer vision in environmental protection. To investigate performance differences among model architectures for small-object waste sorting

this study systematically compared three representative models—the CNN model EfficientNet-B0

the pure attention model ViT-Tiny

and the hybrid model ConvNeXt-Tiny—using a 12-category waste image dataset with particular focus on small objects like "batteries." Experiments conducted under a unified training framework revealed that EfficientNet-B0 achieved the highest accuracy of 95.76%

significantly outperforming ViT-Tiny (75.12%) and ConvNeXt-Tiny (64.93%). A specialized analysis of the small target "battery" further confirmed the CNN model's advantage. Results indicate that in scenarios with small samples and small targets

CNN's inductive bias plays a decisive role in learning discriminative features

providing empirical evidence for selecting models for waste classification.

关键词

Keywords

references

Shuishuo Jun, "Driven Crazy by Garbage Sorting? AI Automatic Sorting Might Be Your Sa vior," Beijing News, 2019.M. Tan and Q. V. Le, "EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks," in International Conference on Machine Learning (ICML) , 2019.

A. Vaswani et al., "Attention is All You Need," in Advances in Neural Information Processing Systems (NeurIPS) , 2017, pp. 5998 – 6008 .

A. Dosovitskiy et al., "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale," in International Conference on Learning Representations (ICLR) , 2021.

Z. Liu et al., "A ConvNet for the 2020s," in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 11976 – 11986 .

T.-Y. Lin et al., "Focal Loss for Dense Object Detection," in Proceedings of the IEEE International Conference on Computer Vision (ICCV) , 2017, pp. 2980 – 2988 .

Q. Qiu et al., "ETHSeg: An Amodel Instance Segmentation Network and a Real-World Dataset for X-Ray Waste Inspection," in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2022.

Chen Jiansong, Cai Yijun, "A Lightweight Object Detection Approach for Waste Sorting Scenarios," Journal of Zhejiang Univers ity (Engineering Science), vol. 58 , no. 1 , pp. 71 – 77 , 2024 .

L. Liu et al., "Ranger: A Synergistic Optimizer Combining RAdam and LookAhead," GitHub Repository , 2020.

Z. Li et al., "Research on Scene Classification in Remote Sensing Images Using ResNet and EfficientNet," Computer Science and Applications, 2024.W. Wang et al., "Co-DETR: Improving DETR-based Detectors with Collaborative Training," in Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , 2023, pp. 15222 – 15231 .

Liu, Wenting, and Lu, Xinming. "Research Progress on Transformers Based on Computer Vision." Computer Engineering and Applications, vol. 58 , no. 6 , pp. 1 – 16 , 2022 .

"Thousands of Teams Compete! First "Qizhi Cup" Algorithm Competition Concludes Successfully, Accelerating AI Application Deployment," Tech News, 2025.