YOLO4SWD: an improved YOLOv11-based method for waste detection and classification
- 1School of Mathematics and Computer Science, Yunnan Minzu University
Abstract
To make sorting more efficient, promote the recycling of resources, and lower the costs of environmental management, it is important to be able to accurately identify waste categories. However, current waste detection methods encounter substantial obstacles, such as inadequate feature extraction for diminutive waste objects, disproportionate contextual information application in intricate backgrounds, and a notable decline in classification accuracy in occlusion situations. To address these limitations, this study introduces an enhanced waste detection model derived from YOLOv11n, termed YOLO4SWD (YOLO for Smart Waste Detection), which integrates three principal innovations. (1)The CoordA module is embedded into the backbone to accurately capture fine local features of tiny waste and boost feature discrimination between waste and low-contrast backgrounds. (2) A lightweight BasicRFB block is integrated into the neck layer to enlarge the model’s effective receptive field.(3) The FSEAM module is adopted in the detection head to preserve the effective features of unoccluded regions and block interference from occluded regions, thereby improving the accuracy of waste detection in scenarios where occlusions are common. The experimental results show that the proposed YOLO4SWD model is much better than the baseline YOLOv11n. The proposed model exhibits robust detection performance, achieving 81.6%, 69.9%, and 62.5% in [email protected], [email protected], and [email protected]:0.95, respectively. This indicates that it performs better in detecting trash in complex environments.
Related
- Cracks-YOLO: an improved YOLOv5 tongue crack detection method based on switchable atrous convolution and transformer
- LHR-YOLO: A SAR small-target ship detection method based on improved YOLO11
- iSL-YOLOv11: global–local attention enhanced small lesion detection in mammography images
- Improved Dissolution Performance of Fenoprofen Calcium Using PEG 6000 Solid Dispersions: Preparation by Fusion Method and Physicochemical Characterization
- STCP-Net: A few-shot learning approach based on spectral-temporal convolution and an improved prototypical network for ECG classification
BibTeX
@article{chen2026yolo4swd,
title = {{YOLO4SWD: an improved YOLOv11-based method for waste detection and classification}},
author = {Xuehao Chen and Fafu Zhang and Wenlin Pan and Yanfang Zhao},
journal = {Scientific Reports},
year = {2026},
month = jul,
doi = {10.1038/s41598-026-62993-3},
url = {https://doi.org/10.1038/s41598-026-62993-3},
note = {CC BY},
}
RIS
TY - JOUR
TI - YOLO4SWD: an improved YOLOv11-based method for waste detection and classification
AU - Xuehao Chen
AU - Fafu Zhang
AU - Wenlin Pan
AU - Yanfang Zhao
PY - 2026
DA - 2026/07/01
JO - Scientific Reports
DO - 10.1038/s41598-026-62993-3
UR - https://doi.org/10.1038/s41598-026-62993-3
AB - To make sorting more efficient, promote the recycling of resources, and lower the costs of environmental management, it is important to be able to accurately identify waste categories. However, current waste detection methods encounter substantial obstacles, such as inadequate feature extraction for diminutive waste objects, disproportionate contextual information application in intricate backgrounds, and a notable decline in classification accuracy in occlusion situations. To address these limitations, this study introduces an enhanced waste detection model derived from YOLOv11n, termed YOLO4SWD (YOLO for Smart Waste Detection), which integrates three principal innovations. (1)The CoordA module is embedded into the backbone to accurately capture fine local features of tiny waste and boost feature discrimination between waste and low-contrast backgrounds. (2) A lightweight BasicRFB block is integrated into the neck layer to enlarge the model’s effective receptive field.(3) The FSEAM module is adopted in the detection head to preserve the effective features of unoccluded regions and block interference from occluded regions, thereby improving the accuracy of waste detection in scenarios where occlusions are common. The experimental results show that the proposed YOLO4SWD model is much better than the baseline YOLOv11n. The proposed model exhibits robust detection performance, achieving 81.6%, 69.9%, and 62.5% in [email protected], [email protected], and [email protected]:0.95, respectively. This indicates that it performs better in detecting trash in complex environments.
SN - 2045-2322
ER -