Plant Phenomics / 2022 / Article / Tab 5

Research Article

PSegNet: Simultaneous Semantic and Instance Segmentation for Point Clouds of Plants

Table 5

Quantitative comparison of semantic segmentation performances of six networks including PSegNet.

MethodsTobaccoTomatoSorghumMean
StemLeafStemLeafStemLeaf

Prec (%)PointNet77.1594.0293.9996.7177.8795.3789.19
PointNet++87.7895.6293.6596.8078.0198.3391.70
ASIS91.6591.9493.5597.1485.4795.1792.49
PlantNet89.4596.8095.9096.3089.0797.4394.16
DGCNN90.5596.4295.2497.8683.9597.3793.57
PSegNet (ours)92.7196.7696.3697.9889.5498.0495.23

Rec (%)PointNet79.2093.3191.8597.6161.4597.8586.88
PointNet++90.8394.0592.4597.3378.6698.2791.93
ASIS83.8596.1192.8795.5181.6597.8891.31
PlantNet86.1292.9795.2498.2386.0698.0792.78
DGCNN85.5597.7694.1598.2778.0598.2092.00
PSegNet (ours)87.4297.7395.0298.5985.6998.6393.85

F1 (%)PointNet78.1693.6692.9197.1668.6996.5987.86
PointNet++89.2894.8393.0597.0778.3398.3091.81
ASIS87.5893.9893.2196.3283.5296.5191.85
PlantNet87.7594.8595.5697.2687.5497.7593.45
DGCNN87.9897.0994.6998.0780.8997.7892.75
PSegNet (ours)89.9997.2495.6898.2987.5798.3394.52

IoU (%)PointNet64.1588.0886.7694.4852.3193.4179.87
PointNet++80.6390.1787.0094.3064.3896.6585.52
ASIS77.9188.6487.2992.9071.7093.2585.28
PlantNet78.1790.2091.5194.6677.8495.5988.00
DGCNN78.5494.3489.9296.2076.9295.6588.60
PSegNet (ours)81.7994.6391.7396.6377.8996.7289.90

The best values are in boldface.