Table 10 Evaluation and comparison of our performance and efficiency in the 1-way 1-shot scenario on the SAA dataset: Training Time and Testing Speed Were Tested on an RTX 2080 Ti

From: Learning discriminative universal background knowledge for few-shot point cloud semantic segmentation of architectural cultural heritage

Method

Memory

# Params

Inference time

mIoU

AttMPTI23

1.62GB

0.35M

120.00 ms

63.95%

QGE29

0.72GB

0.57M

135.90 ms

64.20%

AttProtoNet41

0.06GB

0.35M

10.00 ms

59.55%

QGPA28

0.08GB

2.57M

11.00 ms

66.03%

DPA32

0.09GB

5.08M

12.25 ms

66.15%

Ours

0.25GB

7.25M

60.00 ms

72.41%