CabbageNet: Deep Learning for Cabbage Segmentation
Author Information
Author(s): Tian Yongqiang, Cao Xinyu, Zhang Taihong, Wu Huarui, Zhao Chunjiang, Zhao Yunjie, Chung Yongwha
Primary Institution: Xinjiang Agricultural University
Hypothesis
Can an improved YOLOv8n-seg network enhance cabbage head segmentation accuracy in complex environments for autonomous harvesting?
Conclusion
The CabbageNet model achieves high precision and recall for cabbage segmentation, making it suitable for real-time harvesting applications.
Supporting Evidence
- The model achieved a Mask Precision of 92.2%, Mask Recall of 87.2%, and Mask mAP50 of 95.1%.
- CabbageNet outperformed other models in segmentation accuracy and computational efficiency.
- The model size is only 6.46 MB, making it suitable for real-time applications.
- CabbageNet demonstrated a frame rate of 154 FPS, indicating its capability for real-time processing.
Takeaway
CabbageNet is a smart computer program that helps robots find and pick cabbages accurately, even in messy fields.
Methodology
The study used an improved YOLOv8n-seg network with modifications for better segmentation performance, trained on a dataset of 10,000 cabbage images.
Limitations
The dataset includes only a limited number of cabbage varieties, which may affect the model's generalizability.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website