A comparative study on point cloud down-sampling strategies for deep learning-based crop organ segmentation

Table 11 Quantitative comparison of both semantic segmentation and instance segmentation performances across five down-sampling strategies on the PSegNet network for the Soybean-MVS dataset

PSegNet	Semantic Segmentation (%)				Instance Segmentation (%)				Score	AveDiff (%)
PSegNet	Precision	Recall	F1	IoU	Cov	WCov	mPrec	mRec	Score	AveDiff (%)
FPS	86.75	86.76	86.75	77.65	45.40	51.19	48.76	26.65	15	2.11
RS	89.13	89.01	89.06	80.50	46.40	46.54	47.33	27.31	24	1.43
UVS	85.82	86.81	86.30	76.96	44.01	50.93	49.17	25.14	7	2.70
VFPS	85.13	86.16	85.61	76.08	38.68	45.52	43.67	20.10	0	5.73
3DEPS	88.64	88.31	88.47	79.67	47.99	48.96	50.27	29.64	26	0.60

ISSN: 1746-4811