| Model | bbox AP (val) AP - AP50 - AP75 |
segm AP (val) AP - AP50 - AP75 |
Download |
|---|---|---|---|
| SWA-MaskRCNN-R50-1x-0.02-0.0002-(38.2, 34.7)-(58.8, 55.7)-(41.4, 37.2) | (39.1, +0.9) - (59.6, +0.8) - (42.8, +1.4) | (35.5, +0.8) - (56.7, +1.0) - (37.8, +0.6) | model | config |
| SWA-MaskRCNN-R101-1x-0.02-0.0002-(40.0, 36.1)-(60.5, 57.5)-(44.0, 38.6) | (41.0, +1.0) - (61.4, +0.9) - (45.1, +1.1) | (37.0, +0.9) - (58.5, +1.0) - (39.7, +1.1) | model | config |
| SWA-MaskRCNN-R101-2x-0.02-0.0002-(40.8, 36.6)-(61.0, 57.9)-(44.5, 39.1) | (41.7, +0.9) - (62.1, +1.1) - (45.5, +1.0) | (37.4, +0.8) - (59.1, +1.2) - (40.2, +1.1) | model | config |
| SWA-FasterRCNN-R50-1x-0.02-0.0002-37.4-58.1-40.4 | (38.4, +1.0) - (59.3, +1.2) - (41.6, +1.2) | - | model | config |
| SWA-FasterRCNN-R101-1x-0.02-0.0002-39.4-60.1-43.1 | (40.3, +0.9) - (61.1, +1.0) - (44.0, +0.9) | - | model | config |
| SWA-FasterRCNN-R101-2x-0.02-0.0002-39.8-60.1-43.3 | (40.7, +0.9) - (61.2, +1.1) - (44.5, +1.2) | - | model | config |
| SWA-RetinaNet-R50-1x-0.01-0.0001-36.5-55.4-39.1 | (37.8, +1.3) - (56.9, +1.5) - (40.5, +1.4) | - | model | config |
| SWA-RetinaNet-R101-1x-0.01-0.0001-38.5-57.6-41.0 | (39.7, +1.2) - (58.9, +1.3) - (42.2, +1.2) | - | model | config |
| SWA-RetinaNet-R101-2x-0.01-0.0001-38.9-58.0-41.5 | (40.0, +1.1) - (59.3, +1.3) - (42.9, +1.4) | - | model | config |
| SWA-FCOS-R50-1x-0.01-0.0001-36.6-55.7-38.8 | (38.0, +1.4) - (57.2, +1.5) - (40.2, +1.4) | - | model | config |
| SWA-FCOS-R101-1x-0.01-0.0001-39.2-58.8-42.1 | (40.3, +1.1) - (59.9, +1.1) - (43.0, +0.9) | - | model | config |
| SWA-FCOS-R101-2x-0.01-0.0001-39.1-58.4-41.6 | (40.2, +1.1) - (59.5, +1.1) - (42.7, +1.1) | - | model | config |
| SWA-YOLOv3(320)-D53-273e-0.001-0.00001-27.9-49.2-28.3 | (28.7, +0.8) - (50.1, +0.9) - (29.1, +0.8) | - | model | config |
| SWA-YOLOv3(680)-D53-273e-0.001-0.00001-33.4-56.3-35.2 | (34.2, +0.8) - (57.3, +1.0) - (35.7, +0.5) | - | model | config |
| SWA-VFNet-R50-1x-0.01-0.0001-41.6-59.5-45.0 | (42.8, +1.2) - (60.7, +1.2) - (46.3, +1.3) | - | model | config |
| SWA-VFNet-R101-1x-0.01-0.0001-43.0-60.9-46.8 | (44.3, +1.3) - (62.2, +1.3) - (48.3, +1.5) | - | model | config |
| SWA-VFNet-R101-2x-0.01-0.0001-43.5-61.3-47.2 | (44.5, +1.0) - (62.4, +1.1) - (48.2, +1.0) | - | model | config |
Notes:
- SWA-MaskRCNN-R50-1x-0.02-0.0002-(38.2, 34.7)-(58.8, 55.7)-(41.4, 37.2) means this SWA model is produced based on the pre-trained Mask RCNN model that has a ResNet50 backbone, is trained under 1x schedule with the initial learning rate 0.02 and ending learning rate 0.0002, and achieves 38.2 bbox AP and 34.7 mask AP, 58.8 bbox AP50 and 55.7 mask AP50, 41.4 bbox AP75 and 37.2 mask AP75 on the COCO val2017 respectively. This SWA model acheives 39.1 bbox AP, 59.6 bbox AP50 and 42.8 bbox AP75, and 35.5 mask AP, 56.7 mask AP50 and 37.8 mask AP75, which are higher than the pre-trained model by 0.9 bbox AP, 0.8 bbox AP50 and 1.4 bbox AP75, and 0.8 mask AP, 1.0 mask AP50 and 0.6 mask AP75 respectively. This rule applies to other object detectors.