Persistence diagram reconstruction on ModelNet40 (test)
3,980W2 DistancePointGPT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PointGPTEncoder training protocol=Pretrained-frozen, Feature selection=Last transformer block (L)2026.04 | 3,980 | 12,630 | 10.3 | |
| Point-BERTEncoder training protocol=Pretrained-frozen, Feature selection=Combined features from all transformer blocks (C)2026.04 | 4,192 | 13,090 | 9.9 | |
| Point-BERTEncoder training protocol=Pretrained-frozen, Feature selection=Last transformer block (L)2026.04 | 4,304 | 12,540 | 10.65 | |
| DGCNNEncoder training protocol=Encoders trained end-to-end2026.04 | 4,327 | 12,510 | 9.773 | |
| Point-MAEEncoder training protocol=Pretrained-frozen, Feature selection=Last transformer block (L)2026.04 | 4,568 | 13,110 | 10.29 | |
| PCP-MAEEncoder training protocol=Pretrained-frozen, Feature selection=Combined features from all transformer blocks (C)2026.04 | 4,592 | 12,870 | 9.968 | |
| PCP-MAEEncoder training protocol=Pretrained-frozen, Feature selection=Last transformer block (L)2026.04 | 4,693 | 13,020 | 10.18 | |
| Point-MAEEncoder training protocol=Pretrained-frozen, Feature selection=Combined features from all transformer blocks (C)2026.04 | 4,726 | 13,800 | 11.93 | |
| PointGPTEncoder training protocol=Pretrained-frozen, Feature selection=Combined features from all transformer blocks (C)2026.04 | 4,889 | 12,990 | 10.22 | |
| PointNetEncoder training protocol=Encoders trained end-to-end2026.04 | 5,132 | 12,720 | 9.899 | |
| PointNet++Encoder training protocol=Encoders trained end-to-end2026.04 | 7,886 | 12,980 | 314 |