| _base_ = './lraspp_m-v3-d8_scratch_512x1024_320k_cityscapes.py' | |
| norm_cfg = dict(type='SyncBN', eps=0.001, requires_grad=True) | |
| model = dict( | |
| type='EncoderDecoder', | |
| backbone=dict( | |
| type='MobileNetV3', | |
| arch='small', | |
| out_indices=(0, 1, 12), | |
| norm_cfg=norm_cfg), | |
| decode_head=dict( | |
| type='LRASPPHead', | |
| in_channels=(16, 16, 576), | |
| in_index=(0, 1, 2), | |
| channels=128, | |
| input_transform='multiple_select', | |
| dropout_ratio=0.1, | |
| num_classes=19, | |
| norm_cfg=norm_cfg, | |
| act_cfg=dict(type='ReLU'), | |
| align_corners=False, | |
| loss_decode=dict( | |
| type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0))) | |