_base_ = './deeplabv3plus_r50-d8_512x1024_80k_cityscapes.py' | |
model = dict( | |
pretrained='torchvision://resnet18', | |
backbone=dict(type='ResNet', depth=18), | |
decode_head=dict( | |
c1_in_channels=64, | |
c1_channels=12, | |
in_channels=512, | |
channels=128, | |
), | |
auxiliary_head=dict(in_channels=256, channels=64)) | |