3D Point Cloud Classification On Scanobjectnn

評価指標

Mean Accuracy

OBJ-BG (OA)

OBJ-ONLY (OA)

Overall Accuracy

評価結果

このベンチマークにおける各モデルのパフォーマンス結果

モデル名	Mean Accuracy	OBJ-BG (OA)	OBJ-ONLY (OA)	Overall Accuracy	Paper Title	Repository
point2vec	86.0	91.2	90.4	87.5	Point2Vec for Self-Supervised Representation Learning on Point Clouds
ULIP-2 + Point-BERT	-	-	-	89.0	ULIP-2: Towards Scalable Multimodal Pre-training for 3D Understanding
Point-BERT	-	87.43	88.12	83.1	Point-BERT: Pre-training 3D Point Cloud Transformers with Masked Point Modeling
Point-PN	-	-	-	87.1	Parameter is Not All You Need: Starting from Non-Parametric Networks for 3D Point Cloud Analysis
ULIP-2 + PointNeXt (no voting)	90.3	-	-	90.8	ULIP-2: Towards Scalable Multimodal Pre-training for 3D Understanding
PointNet	63.4	-	-	68.2	PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation
PatchAugment	79.7	-	-	81.0	PatchAugment: Local Neighborhood Augmentation in Point Cloud Classification
Mamba3D (no voting)	-	92.94	92.08	91.81	Mamba3D: Enhancing Local Features for 3D Point Cloud Analysis via State Space Model
Point-TnT	81.0	-	-	83.5	Points to Patches: Enabling the Use of Self-Attention for 3D Shape Recognition
Point-JEPA	-	92.9±0.4	-	-	Point-JEPA: A Joint Embedding Predictive Architecture for Self-Supervised Learning on Point Cloud	-
PointNeXt+HyCoRe	87.0	-	-	88.3	Rethinking the compositionality of point clouds through regularization in the hyperbolic space
RepSurf-U	-	-	-	84.6	Surface Representation for Point Clouds
ACT	-	-	-	89.17	Autoencoders as Cross-Modal Teachers: Can Pretrained 2D Image Transformers Help 3D Representation Learning?
ReCon (no voting)	-	95.18	93.29	90.63	Contrast with Reconstruct: Contrastive 3D Representation Learning Guided by Generative Pretraining
SimpleView	-	-	-	80.5	Revisiting Point Cloud Classification with a Simple and Effective Baseline
I2P-MAE (no voting)	-	94.15	91.57	90.11	Learning 3D Representations from 2D Pre-trained Models via Image-to-Point Masked Autoencoders
KPConvX-L	88.1	-	-	89.3	KPConvX: Modernizing Kernel Point Convolution with Kernel Attention	-
ULIP + PointBERT	-	-	-	86.4	ULIP: Learning a Unified Representation of Language, Images, and Point Clouds for 3D Understanding
RepSurf-U (2x)	-	-	-	86.0	Surface Representation for Point Clouds
PointNeXt	86.8	-	-	88.2	PointNeXt: Revisiting PointNet++ with Improved Training and Scaling Strategies

0 of 75 row(s) selected.