Image Segmentation
Transformers
PyTorch
upernet
Inference Endpoints
File size: 716 Bytes
b13b124
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
_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)))