| _base_ = './rpn_r50_fpn_1x_coco.py' | |
| # learning policy | |
| max_epochs = 24 | |
| train_cfg = dict( | |
| type='EpochBasedTrainLoop', max_epochs=max_epochs, val_interval=1) | |
| param_scheduler = [ | |
| dict( | |
| type='LinearLR', start_factor=0.001, by_epoch=False, begin=0, end=500), | |
| dict( | |
| type='MultiStepLR', | |
| begin=0, | |
| end=max_epochs, | |
| by_epoch=True, | |
| milestones=[16, 22], | |
| gamma=0.1) | |
| ] | |