Vanishing Point Detection on ScanNet (test)
64.4AA@10°NeurVPS
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| NeurVPSParams=22 M, FPS=0.52022.03 | 64.4 | 24 | 41.8 | |
| OursParams=7 M, FPS=5.52022.03 | 63.7 | 24.8 | 42.1 | |
| Ours*Params=5 M, FPS=23, multi-scale sampling=true2022.03 | 62.4 | 22.9 | 39.8 | |
| Ours†Params=7 M, FPS=5.5, unknown focal length=true2022.03 | 59.9 | 22.2 | 38.8 | |
| J-Linkage†FPS=1, unknown focal length=true2022.03 | 43 | 15.7 | 27.3 | |
| Deep Vanishing Point Detection with Geometric PriorsTraining Dataset=SU3 [70]2022.03 | 39.5 | 15.2 | 25.9 | |
| Quasi-VPFPS=292022.03 | 39.4 | 14.7 | 25.3 | |
| CONSAC†Params=0.2 M, FPS=3, unknown focal length=true2022.03 | 36 | 15.8 | 24.6 | |
| NeurVPSTraining Dataset=SU3 [70]2022.03 | 35.5 | 11.1 | 20.3 | |
| Contrario-VP†FPS=0.6, unknown focal length=true2022.03 | 35.3 | 12 | 21.4 | |
| Deep Vanishing Point Detection with Geometric PriorsTraining Dataset=NYU Depth [45]2022.03 | 33.6 | — | — | |
| CONSACTraining Dataset=SU3 [70]2022.03 | 27.2 | 10.1 | 17.3 | |
| CONSACTraining Dataset=NYU Depth [45]2022.03 | 20.3 | — | — |