MMDetection
MMDetection is an open source object detection toolbox based on PyTorch. It consists of:
- Training recipes for object detection and instance segmentation.
- 360+ pre-trained models to use for fine-tuning (or training afresh).
- Dataset support for popular vision datasets such as COCO, Cityscapes, LVIS and PASCAL VOC.
Models
Viewing Models for Object Detection:
| MODEL | BOX AP | INFERENCE TIME (S/IM) | PAPER | YEAR | |
|---|---|---|---|---|---|
|
|
SCNet |
47.5
|
0.2041
|
||
|
|
Sparse R-CNN |
46.2
|
|
||
|
|
VFNet |
50.4
|
|
||
|
|
PAA |
45.1
|
|
||
|
|
Generalized Focal Loss |
48.1
|
0.0935
|
||
|
|
DetectoRS |
49.1
|
|
||
|
|
DETR |
40.1
|
|
||
|
|
GRoIE |
42.2
|
|
||
|
|
ResNeSt |
47.7
|
|
||
|
|
Dynamic R-CNN |
38.9
|
|
||
|
|
RegNet |
43.1
|
|
||
|
|
CentripetalNet |
44.8
|
0.2703
|
||
|
|
PointRend |
41.0
|
|
||
|
|
SABL |
43.6
|
|
||
|
|
ATSS |
41.5
|
0.0813
|
||
|
|
FreeAnchor |
41.9
|
0.0901
|
||
|
|
InstaBoost |
44.7
|
|
||
|
|
Cascade R-CNN |
44.7
|
|
||
|
|
Cascade Mask R-CNN |
45.6
|
|
||
|
|
NAS-FCOS |
39.4
|
|
||
|
|
CARAFE |
39.3
|
0.0714
|
||
|
|
RepPoints |
44.2
|
0.1075
|
||
|
|
GCNet |
46.4
|
|
||
|
|
NAS-FPN |
40.5
|
0.0435
|
||
|
|
PISA |
41.9
|
|
||
|
|
FoveaBox |
42.0
|
0.068
|
||
|
|
Libra R-CNN |
42.7
|
0.1176
|
||
|
|
Res2Net |
47.5
|
|
||
|
|
FCOS |
42.6
|
0.1031
|
||
|
|
GroupNorm + WS |
43.1
|
|
||
|
|
FSAF |
42.4
|
0.1786
|
||
|
|
Mask Scoring R-CNN |
43.0
|
0.125
|
||
|
|
HRNet |
46.4
|
|
||
|
|
HTC |
50.4
|
|
||
|
|
Guided Anchoring |
43.9
|
0.137
|
||
|
|
TridentNet |
40.3
|
|
||
|
|
Grid R-CNN |
43.0
|
0.1299
|
||
|
|
DCN |
47.3
|
|
||
|
|
GHM |
41.4
|
0.1923
|
||
|
|
CornerNet |
41.2
|
0.2381
|
||
|
|
YOLOv3 |
33.4
|
0.0208
|
||
|
|
GroupNorm |
42.1
|
|
||
|
|
PAFPN |
37.5
|
0.0581
|
||
|
|
RetinaNet |
41.0
|
0.1149
|
||
|
|
Mask R-CNN |
44.0
|
|
||
|
|
SSD |
29.4
|
0.0326
|
||
|
|
Faster R-CNN |
42.1
|
0.1064
|
||
|
|
Fast R-CNN |
39.9
|
|
Models
Viewing Models for Instance Segmentation:
| MODEL | MASK AP | INFERENCE TIME (S/IM) | PAPER | YEAR | |
|---|---|---|---|---|---|
|
|
SCNet |
42.3
|
0.2041
|
||
|
|
DetectoRS |
42.6
|
|
||
|
|
GRoIE |
37.8
|
|
||
|
|
ResNeSt |
41.4
|
|
||
|
|
RegNet |
38.0
|
|
||
|
|
PointRend |
38.0
|
|
||
|
|
InstaBoost |
39.7
|
|
||
|
|
Cascade Mask R-CNN |
39.5
|
|
||
|
|
CARAFE |
38.6
|
0.0606
|
||
|
|
GCNet |
40.1
|
|
||
|
|
Res2Net |
41.6
|
|
||
|
|
GroupNorm + WS |
38.6
|
|
||
|
|
Mask Scoring R-CNN |
39.5
|
0.125
|
||
|
|
HRNet |
40.8
|
|
||
|
|
HTC |
43.8
|
|
||
|
|
DCN |
41.1
|
|
||
|
|
GroupNorm |
38.0
|
|
||
|
|
Mask R-CNN |
39.3
|
|
