Comparing Mask R-CNN and YOLOv8 for Stomatal Pore Analysis
Author Information
Author(s): Thai Thanh Tuan, Ku Ki-Bon, Le Anh Tuan, Oh San Su Min, Phan Ngo Hoang, Kim In-Jung, Chung Yong Suk
Primary Institution: Jeju National University
Hypothesis
This study aims to compare the effectiveness of Mask R-CNN and YOLOv8 in segmenting stomatal pores from plant images.
Conclusion
YOLOv8 outperformed Mask R-CNN in accurately calculating stomatal pore dimensions and handling image imperfections.
Supporting Evidence
- YOLOv8 achieved a mean average precision of 0.98164, surpassing Mask R-CNN's 0.97920.
- YOLOv8 demonstrated lower Mean Absolute Error values for width, length, and area estimations compared to Mask R-CNN.
- Both models were evaluated against ground truth annotations to assess their segmentation accuracy.
Takeaway
The study tested two computer programs to see which one is better at finding tiny holes in plant leaves called stomata, and one program did a better job.
Methodology
The study used a novel dataset for stomatal pore segmentation and compared the performance of Mask R-CNN and YOLOv8 through rigorous evaluation metrics.
Limitations
The study did not modify the fundamental structures of the algorithms, which may limit the exploration of potential improvements.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website