Using Neural Networks for 3D Segmentation of Blueberries
Author Information
Author(s): Mullins Connor C., Esau Travis J., Zaman Qamar U., Al-Mallahi Ahmad A., Farooque Aitazaz A.
Primary Institution: Dalhousie University
Hypothesis
Can 2D neural network frameworks effectively segment 3D images of harvested wild blueberries?
Conclusion
The YOLOv8n-seg model is recommended for real-time 3D segmentation in precision agriculture due to its superior performance and processing speed.
Supporting Evidence
- YOLOv8 models showed an average processing time of 18.10 ms, significantly faster than Detectron2's 57 ms.
- Depth map segmentation achieved a mean intersection over union (IoU) of 0.944 and a Matthew’s correlation coefficient (MCC) of 0.957.
- Statistical analyses confirmed the superior performance of depth map models over RGB models.
Takeaway
This study shows that using special computer programs can help farmers better see and measure blueberries in 3D, making it easier to know how much they have.
Methodology
The study used YOLOv8 and Detectron2 models to analyze depth maps and RGB images of wild blueberries, comparing their segmentation performance.
Limitations
The study primarily focused on wild blueberries and may not generalize to other crops or conditions.
Statistical Information
P-Value
p < 0.001
Statistical Significance
p < 0.001
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website