Automated detection of defective coffee beans based on improved YOLOv10 framework.

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Manual coffee quality assessment is labor-intensive and inconsistent.
• Automated defect detection is crucial for the global coffee economy.

Purpose of the Study:

• To develop an improved YOLOv10-based framework for automated detection of defective green coffee beans.
• To achieve a balance between high-precision localization and real-time edge deployment.

Main Methods:

• Integrated Depthwise Separable Convolution (DSConv) with distribution shift for feature extraction.
• Employed Spatial Pyramid Pooling Fast with Attention (SPPF_Attention) for multi-scale feature refinement.
• Utilized Partial Convolution (PConv) operators in the detection head for computational efficiency and robustness.

Main Results:

• Achieved state-of-the-art performance with 99.2% mAP@50, 98.5% precision, and 98.8% recall.
• Demonstrated rapid inference latency of 2.0 ms.
• Reduced model parameters by 21.6% and model size by 12.4% compared to YOLOv11 baseline.

Conclusions:

• The proposed framework is a highly viable and scalable solution for resource-constrained industrial sorting facilities.
• The model exhibits exceptional generalization capability across diverse processing states.
• Architectural innovations contribute to improved accuracy, efficiency, and robustness in defect detection.