3D Point Cloud Classification On Scanobjectnn
Metrics
Mean Accuracy
OBJ-BG (OA)
OBJ-ONLY (OA)
Overall Accuracy
Results
Performance results of various models on this benchmark
Comparison Table
| Model Name | Mean Accuracy | OBJ-BG (OA) | OBJ-ONLY (OA) | Overall Accuracy |
|---|---|---|---|---|
| point2vec-for-self-supervised-representation | 86.0 | 91.2 | 90.4 | 87.5 |
| ulip-2-towards-scalable-multimodal-pre | - | - | - | 89.0 |
| point-bert-pre-training-3d-point-cloud | - | 87.43 | 88.12 | 83.1 |
| parameter-is-not-all-you-need-starting-from | - | - | - | 87.1 |
| ulip-2-towards-scalable-multimodal-pre | 90.3 | - | - | 90.8 |
| pointnet-deep-learning-on-point-sets-for-3d | 63.4 | - | - | 68.2 |
| patchaugment-local-neighborhood-augmentation | 79.7 | - | - | 81.0 |
| mamba3d-enhancing-local-features-for-3d-point | - | 92.94 | 92.08 | 91.81 |
| points-to-patches-enabling-the-use-of-self | 81.0 | - | - | 83.5 |
| point-jepa-a-joint-embedding-predictive | - | 92.9±0.4 | - | - |
| rethinking-the-compositionality-of-point | 87.0 | - | - | 88.3 |
| surface-representation-for-point-clouds | - | - | - | 84.6 |
| autoencoders-as-cross-modal-teachers-can | - | - | - | 89.17 |
| contrast-with-reconstruct-contrastive-3d | - | 95.18 | 93.29 | 90.63 |
| revisiting-point-cloud-classification-with-a | - | - | - | 80.5 |
| learning-3d-representations-from-2d-pre | - | 94.15 | 91.57 | 90.11 |
| kpconvx-modernizing-kernel-point-convolution | 88.1 | - | - | 89.3 |
| ulip-learning-unified-representation-of | - | - | - | 86.4 |
| surface-representation-for-point-clouds | - | - | - | 86.0 |
| pointnext-revisiting-pointnet-with-improved | 86.8 | - | - | 88.2 |
| beyond-first-impressions-integrating-joint | 88.7 | - | - | 89.5 |
| dynamic-local-geometry-capture-in-3d | - | - | - | 82.0 |
| mvtn-multi-view-transformation-network-for-3d | - | 92.6 | 92.3 | 82.8 |
| p2p-tuning-pre-trained-image-models-for-point | - | - | - | 89.3 |
| pointgpt-auto-regressively-generative-pre-1 | - | 97.2 | 96.6 | 93.4 |
| autoencoders-as-cross-modal-teachers-can | - | 93.29 | 91.91 | 88.21 |
| sagemix-saliency-guided-mixup-for-point | - | - | - | 83.6 |
| modelnet-o-a-large-scale-synthetic-dataset | - | - | - | 86.6 |
| a-deep-dive-into-explainable-self-supervised | - | 90.88 | 90.02 | - |
| ulip-2-towards-scalable-multimodal-pre | 91.2 | - | - | 91.5 |
| point-cloud-classification-using-content | 88.5 | - | - | 90.3 |
| advanced-feature-learning-on-point-clouds | 86.2 | - | - | 87.2 |
| gpsformer-a-global-perception-and-local | 92.51 | - | - | 93.30 |
| point-m2ae-multi-scale-masked-autoencoders | - | 91.22 | 88.81 | 86.43 |
| positional-prompt-tuning-for-efficient-3d | - | 95.01 | 93.28 | 89.52 |
| beyond-local-patches-preserving-global-local | 87.2 | - | - | 89.0 |
| 3d-jepa-a-joint-embedding-predictive | - | 93.63 | 94.49 | 89.52 |
| masked-autoencoders-for-point-cloud-self | - | 90.02 | 88.29 | 85.2 |
| 0-4-dualities | - | - | - | 88.4 |
| sagemix-saliency-guided-mixup-for-point | - | - | - | 83.7 |
| rethinking-network-design-and-local-geometry-1 | 81.8 | - | - | 83.8 |
| instance-aware-dynamic-prompt-tuning-for-pre | - | 93.12 | - | 88.51 |
| gpsformer-a-global-perception-and-local | 93.8 | - | - | 95.4 |
| let-images-give-you-more-point-cloud-cross | 84.8 | - | - | 86.7 |
| app-net-auxiliary-point-based-push-and-pull | - | - | - | 84.1 |
| self-positioning-point-based-transformer-for | 86.8 | - | - | 88.6 |
| rethinking-masked-representation-learning-for | - | 92.9 | 92.3 | 89.0 |
| mamba3d-enhancing-local-features-for-3d-point | - | 94.49 | 92.43 | 92.64 |
| pra-net-point-relation-aware-network-for-3d | 79.1 | - | - | 82.1 |
| shapellm-universal-3d-object-understanding | - | 98.80 | 97.59 | 95.25 |
| pcp-mae-learning-to-predict-centers-for-point | - | 95.52 | 94.32 | 90.35 |
| point-cloud-classification-using-content | 86.0 | - | - | 88.0 |
| deltaconv-anisotropic-point-cloud-learning | - | - | - | 84.7 |
| point-lgmask-local-and-global-contexts | - | 89.8 | 89.3 | 85.3 |
| dense-resolution-network-for-point-cloud | 78.0 | - | - | 80.3 |
| towards-compact-3d-representations-via-point | - | 95.18 | 93.29 | 90.22 |
| omnivec-learning-robust-representations-with | - | - | - | 96.1 |
| pointcnn-convolution-on-mathcalx-transformed | 75.1 | 86.1 | 85.5 | 78.5 |
| regress-before-construct-regress-autoencoder | - | 95.53 | 93.63 | 90.28 |
| take-a-photo-3d-to-2d-generative-pre-training | - | - | - | 88.5 |
| pointnet-deep-hierarchical-feature-learning | 75.4 | 82.3 | 84.3 | 77.9 |
| dynamic-graph-cnn-for-learning-on-point | 73.6 | 82.8 | 86.2 | 78.1 |
| omnivec2-a-novel-transformer-based-network | - | - | - | 97.2 |
| dualmlp-a-two-stream-fusion-model-for-3d | - | - | - | 86.4 |
| local-neighborhood-features-for-3d-1 | 87.4 | - | - | 88.6 |
| spidercnn-deep-learning-on-point-sets-with | 69.8 | - | - | 73.7 |
| simpleview-neighborhood-views-for-point-cloud | - | - | - | 84.8 |
| geometric-feedback-network-for-point-cloud | 77.8 | - | - | 80.5 |
| point-is-a-vector-a-feature-representation-in | 86.2 | - | - | 87.8 |
| parameter-efficient-fine-tuning-in-spectral | - | 99.48 | 97.76 | 96.18 |
| contrast-with-reconstruct-contrastive-3d | - | 95.35 | 93.80 | 91.26 |
| rethinking-network-design-and-local-geometry-1 | 84.4 | - | - | 85.7 |
| decoupled-local-aggregation-for-point-cloud | 89.3 | - | - | 90.4 |
| ulip-learning-unified-representation-of | 88.5 | - | - | 89.4 |
| ulip-learning-unified-representation-of | 88.6 | - | - | 89.7 |