Class-agnostic 2D Object Detection on Waymo Open Dataset (WOD) (val)
54.4AP50Supervised Baseline
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Supervised Baselineannotation setting=fully manual annotation, #images=158k, #bboxes=1137k, network weights initialized from=ImageNet2022.10 | 54.4 | 20.5 | 72.4 | 90.9 | 62.8 | 35.5 | 80.8 | 94 | |
| Supervised Baselineannotation setting=fully manual annotation, #images=158k, #bboxes=1137k, network weights initialized from=scratch2022.10 | 52.5 | 23.5 | 67.6 | 86.3 | 62.3 | 34.9 | 80 | 93.3 | |
| ClusterNetannotation setting=unsupervised, #images=158k, #bboxes=0, network weights initialized from=scratch, ground truth scene flow=true2022.10 | 51.8 | 21.3 | 70.2 | 89.5 | 60.8 | 30.2 | 81.2 | 94.8 | |
| ClusterNetannotation setting=unsupervised, #images=158k, #bboxes=0, network weights initialized from=scratch2022.10 | 43.2 | 18.4 | 56.5 | 81.8 | 55.4 | 26.7 | 71.9 | 93.1 | |
| Supervised Baselineannotation setting=10% manual annotation, #images=15k, #bboxes=127k, network weights initialized from=ImageNet2022.10 | 33.8 | 5.5 | 45.3 | 74.9 | 36.1 | 9.7 | 48.6 | 76.7 | |
| Supervised Baselineannotation setting=10% manual annotation, #images=15k, #bboxes=127k, network weights initialized from=scratch2022.10 | 31.6 | 5.7 | 42.5 | 72.2 | 35.9 | 8.6 | 47.7 | 75.3 | |
| LOSTannotation setting=unsupervised, #images=158k, #bboxes=0, network weights initialized from=ImageNet2022.10 | 1.9 | 0 | 1 | 7.6 | 5 | 0 | 0.4 | 27.9 | |
| FreeSoloannotation setting=unsupervised, #images=158k, #bboxes=0, network weights initialized from=ImageNet2022.10 | 1 | 0.2 | 1 | 1.9 | 2.2 | 0 | 0.1 | 12.7 | |
| Felzenszwalbannotation setting=unsupervised, #images=158k, #bboxes=0, network weights initialized from=ImageNet2022.10 | 0.4 | 0 | 0.5 | 1.1 | 11.1 | 0.6 | 14.5 | 30.7 |