notesum.ai

Published at November 22

cs.CV

cs.AI

Released Date: November 22, 2024

Authors: Yi Wang¹, Jiaze Wang², Ziyu Guo², Renrui Zhang², Donghao Zhou², Guangyong Chen³, Anfeng Liu¹, Pheng-Ann Heng²

Aff.: ¹Central South University; ²The Chinese University of Hong Kong; ³Zhejiang Lab

Supervised Learning Only
Methods	Reference	ScanObjectNN			ModelNet40
Methods	Reference	OBJ-BG	OBJ-ONLY	PB-T50-RS	ModelNet40
PointNet [33]	CVPR 17	73.3	79.2	68.0	89.0
PointNet++ [34]	NeurIPS 17	82.3	84.3	77.9	90.2
PointCNN [22]	NeurIPS 18	86.1	85.5	78.5	91.7
DGCNN [47]	TOG 19	82.8	86.2	78.1	92.0
PRANet [8]	TIP 21	-	-	81.0	92.9
MVTN [15]	ICCV 21	-	-	82.8	93.8
PointNeXt [37]	NeurIPS 22	-	-	87.7	92.9
PointMLP [28]	ICLR 22	-	-	85.4	94.1
RepSurf-U [40]	CVPR 22	-	-	84.3	93.8
ADS [18]	ICCV 23	-	-	87.5	94.0
with Self-Supervised Representation Learning
MaskPoint [25]	CVPR 22	89.3	88.1	84.3	92.6
Point-M2AE [59]	NeurIPS 22	91.2	88.8	86.4	93.4
PointDif [64]	CVPR 24	93.3	91.9	87.6	-
GPM [23]	CVPR 24	90.2	90.0	84.8	93.3
PointMamba [24]	NeurIPS 24	90.7	88.5	84.9	93.6
Point-BERT [57]	CVPR 22	87.4	88.1	83.1	92.7
+PointACL	-	89.5 (+2.1)	88.6 (+0.5)	84.5 (+1.4)	93.1 (+0.4)
Point-MAE [32]	ECCV 22	90.0	88.3	85.2	93.2
+PointACL	-	90.9 (+0.9)	88.8 (+0.5)	85.4 (+0.2)	93.7 (+0.5)
PointGPT-S [5]	NeurIPS 23	91.6	90.0	86.9	93.3
+PointACL	-	92.3 (+0.7)	91.6 (+1.6)	87.1 (+0.2)	93.5 (+0.2)
Point-MAE* [32]	ECCV 22	92.8	91.2	89.0	93.8
+PointACL*	-	93.1 (+0.3)	91.7 (+0.5)	89.2 (+0.2)	94.1 (+0.3)
PointGPT-S* [5]	NeurIPS 23	93.4	92.4	89.2	94.0
+PointACL*	-	94.5 (+1.1)	93.5 (+1.1)	89.9 (+0.7)	94.1 (+0.1)