A cotton organ segmentation method with phenotypic measurements from a point cloud using a transformer

Table 4 Segmentation results from three deep learning networks on the test set

Metric	Organ	PointNet	PointNet + +	TPointNetPlus
Accuracy (%)	Cotton leaf	80.04	92.12	98.39
	Cotton boll	30	86.63	93.59
	Cotton stem	37.87	72.12	83.03
	Overall	61.4	90.53	95.3
F1 score (%)	Cotton leaf	88.57	98.65	99.18
	Cotton boll	20	85.03	96.32
	Cotton stem	52.36	83.33	90.66
	Overall	57.23	90.1	95.22
mIoU (%)	Cotton leaf	61.3	90.4	95.11
	Cotton boll	14	68.83	84.11
	Cotton stem	19.06	64.87	80.37
	Overall	26.79	74.69	86.53

The segmentation results were evaluated in terms of four dimensions: accuracy (higher is better), F1 score (higher is better), and mIoU
The bold numbers indicate the best results in the respective category

ISSN: 1746-4811