Research Article
Refined Voting and Scene Feature Fusion for 3D Object Detection in Point Clouds
Table 3
3D object detection results on the ScanNet V2 [
16] validation set.
| ScanNet V2 | Input | [email protected] | [email protected] |
| DSS [1] | Geo + RGB | 15.2 | 6.8 | F-PointNet [48] | Geo + RGB | 19.8 | 10.8 | GSPN [49] | Geo + RGB | 30.6 | 17.7 | 3DSIS [50] | Geo + 5 views | 40.2 | 22.5 | VoteNet [12] | Geo only | 58.6 | 33.5 | HGNet [14] | Geo only | 61.3 | 34.4 | MLCVNet [13] | Geo only | 64.7 | 42.1 | H3DNet (1BB) [34] | Geo only | 64.4 | 43.4 | Pointformer [33] | Geo only | 64.1 | — | 3DETR [44] | Geo only | 65.0 | 47.0 | BRNet [15] | Geo only | 66.1 | 50.9 | DisARM [52] | Geo only | 64.2 | 46.5 | RSFF-Net (ours) | Geo only | 65.9 | 44.9 |
|
|
Bold denotes the overall best result.
|