All models and benchmarks results are recorded below.
Remarks:
- If not specified, the models are trained 200 epochs.
In following tables, we only displayed ImageNet Linear Evaluation, COCO17 Object Detection and PASCAL VOC12 Aug Segmentation, you can click the model name above to get the comprehensive benchmark results.
If not specified, we use linear evaluation setting from MoCo. Or the settings is mentioned in Remark.
Algorithm | Config | Remarks | Top-1 (%) |
---|---|---|---|
BYOL | byol_resnet50_8xb32-accum16-coslr-200e_in1k | 67.68 | |
DeepCLuster | deepcluster-sobel_resnet50_8xb64-steplr-200e_in1k.py | 46.92 | |
DenseCL | densecl_resnet50_8xb32-coslr-200e_in1k | 63.34 | |
MoCo v2 | mocov2_resnet50_8xb32-coslr-200e_in1k | 67.56 | |
NPID | npid_resnet50_8xb32-steplr-200e_in1k | 58.16 | |
ODC | odc_resnet50_8xb64-steplr-440e_in1k | 53.42 | |
Relative Location | relative-loc_resnet50_8xb64-steplr-70e_in1k | 39.65 | |
Rotation Prediction | rotation-pred_resnet50_8xb16-steplr-70e_in1k | 44.35 | |
SimCLR | simclr_resnet50_8xb32-coslr-200e_in1k | 58.92 | |
SimSiam | simsiam_resnet50_8xb32-coslr-100e_in1k | SimSiam paper setting | 67.88 |
simsiam_resnet50_8xb32-coslr-200e_in1k | SimSiam paper setting | 69.80 | |
SwAV | swav_resnet50_8xb32-mcrop-2-6-coslr-200e_in1k-224-96 | SwAV paper setting | 70.55 |
In COCO17 Object detection task, we choose the evluation protocol from MoCo, with Mask-RCNN architecture, the results below are trained with the same config.
Algorithm | Config | mAP (Box) | mAP (Mask) |
---|---|---|---|
BYOL | byol_resnet50_8xb32-accum16-coslr-200e_in1k | 40.9 | 36.8 |
DenseCL | densecl_resnet50_8xb32-coslr-200e_in1k | ||
MoCo v2 | mocov2_resnet50_8xb32-coslr-200e_in1k | 40.2 | 36.1 |
NPID | npid_resnet50_8xb32-steplr-200e_in1k | ||
Relative Location | relative-loc_resnet50_8xb64-steplr-70e_in1k | 37.5 | 33.7 |
Rotation Prediction | rotation-pred_resnet50_8xb16-steplr-70e_in1k | 37.9 | 34.2 |
SimCLR | simclr_resnet50_8xb32-coslr-200e_in1k | 38.7 | 34.9 |
SimSiam | simsiam_resnet50_8xb32-coslr-100e_in1k | 38.3 | 34.4 |
simsiam_resnet50_8xb32-coslr-200e_in1k | 38.8 | 34.9 | |
SwAV | swav_resnet50_8xb32-mcrop-2-6-coslr-200e_in1k-224-96 | 40.2 | 36.3 |
In Pascal VOC12 Aug Segmentation task, we choose the evluation protocol from MMSeg, with FCN architecture, the results below are trained with the same config.
Algorithm | Config | mIOU |
---|---|---|
BYOL | byol_resnet50_8xb32-accum16-coslr-200e_in1k | 67.16 |
DeepCLuster | deepcluster-sobel_resnet50_8xb64-steplr-200e_in1k | 59.69 |
DenseCL | densecl_resnet50_8xb32-coslr-200e_in1k | 69.47 |
MoCo v2 | mocov2_resnet50_8xb32-coslr-200e_in1k | 67.55 |
NPID | npid_resnet50_8xb32-steplr-200e_in1k | 65.45 |
ODC | odc_resnet50_8xb64-steplr-440e_in1k | 54.76 |
Relative Location | relative-loc_resnet50_8xb64-steplr-70e_in1k | 63.49 |
Rotation Prediction | rotation-pred_resnet50_8xb16-steplr-70e_in1k | 64.31 |
SimCLR | simclr_resnet50_8xb32-coslr-200e_in1k | 64.03 |
SimSiam | simsiam_resnet50_8xb32-coslr-100e_in1k | 46.11 |
simsiam_resnet50_8xb32-coslr-200e_in1k | 46.27 | |
SwAV | swav_resnet50_8xb32-mcrop-2-6-coslr-200e_in1k-224-96 | 63.73 |