Skip to content

Commit 3e3ed9a

Browse files
MengzhangLIMengzhangLI
authored
[Fix] Fix bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes.py (open-mmlab#1901)
1 parent 5b2f19a commit 3e3ed9a

File tree

3 files changed

+65
-5
lines changed

3 files changed

+65
-5
lines changed

‎configs/bisenetv2/README.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -42,7 +42,7 @@ The low-level details and high-level semantics are both essential to the semanti
4242
| Method | Backbone | Crop Size | Lr schd | Mem (GB) | Inf time (fps) | mIoU | mIoU(ms+flip) | config | download |
4343
| ---------------- | --------- | --------- | ------: | -------- | -------------- | ----: | ------------: | ---------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
4444
| BiSeNetV2 | BiSeNetV2 | 1024x1024 | 160000 | 7.64 | 31.77 | 73.21 | 75.74 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/bisenetv2/bisenetv2_fcn_4x4_1024x1024_160k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/bisenetv2/bisenetv2_fcn_4x4_1024x1024_160k_cityscapes/bisenetv2_fcn_4x4_1024x1024_160k_cityscapes_20210902_015551-bcf10f09.pth) \| [log](https://download.openmmlab.com/mmsegmentation/v0.5/bisenetv2/bisenetv2_fcn_4x4_1024x1024_160k_cityscapes/bisenetv2_fcn_4x4_1024x1024_160k_cityscapes_20210902_015551.log.json) |
45-
| BiSeNetV2 (OHEM) | BiSeNetV2 | 1024x1024 | 160000 | 7.64 | - | 73.57 | 75.80 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/bisenetv2/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/bisenetv2/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes_20210902_112947-5f8103b4.pth) \| [log](https://download.openmmlab.com/mmsegmentation/v0.5/bisenetv2/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes_20210902_112947.log.json) |
45+
| BiSeNetV2 (OHEM) | BiSeNetV2 | 1024x1024 | 160000 | 7.64 | - | 75.30 | 77.06 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/bisenetv2/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/bisenetv2/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes_20220808_172324-8bf0aaba.pth) \| [log](https://download.openmmlab.com/mmsegmentation/v0.5/bisenetv2/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes_20220808_172324.log.json) |
4646
| BiSeNetV2 (4x8) | BiSeNetV2 | 1024x1024 | 160000 | 15.05 | - | 75.76 | 77.79 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/bisenetv2/bisenetv2_fcn_4x8_1024x1024_160k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/bisenetv2/bisenetv2_fcn_4x8_1024x1024_160k_cityscapes/bisenetv2_fcn_4x8_1024x1024_160k_cityscapes_20210903_000032-e1a2eed6.pth) \| [log](https://download.openmmlab.com/mmsegmentation/v0.5/bisenetv2/bisenetv2_fcn_4x8_1024x1024_160k_cityscapes/bisenetv2_fcn_4x8_1024x1024_160k_cityscapes_20210903_000032.log.json) |
4747
| BiSeNetV2 (FP16) | BiSeNetV2 | 1024x1024 | 160000 | 5.77 | 36.65 | 73.07 | 75.13 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/bisenetv2/bisenetv2_fcn_fp16_4x4_1024x1024_160k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/bisenetv2/bisenetv2_fcn_fp16_4x4_1024x1024_160k_cityscapes/bisenetv2_fcn_fp16_4x4_1024x1024_160k_cityscapes_20210902_045942-b979777b.pth) \| [log](https://download.openmmlab.com/mmsegmentation/v0.5/bisenetv2/bisenetv2_fcn_fp16_4x4_1024x1024_160k_cityscapes/bisenetv2_fcn_fp16_4x4_1024x1024_160k_cityscapes_20210902_045942.log.json) |
4848

‎configs/bisenetv2/bisenetv2.yml‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -45,10 +45,10 @@ Models:
4545
- Task: Semantic Segmentation
4646
Dataset: Cityscapes
4747
Metrics:
48-
mIoU: 73.57
49-
mIoU(ms+flip): 75.8
48+
mIoU: 75.3
49+
mIoU(ms+flip): 77.06
5050
Config: configs/bisenetv2/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes.py
51-
Weights: https://download.openmmlab.com/mmsegmentation/v0.5/bisenetv2/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes_20210902_112947-5f8103b4.pth
51+
Weights: https://download.openmmlab.com/mmsegmentation/v0.5/bisenetv2/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes_20220808_172324-8bf0aaba.pth
5252
- Name: bisenetv2_fcn_4x8_1024x1024_160k_cityscapes
5353
In Collection: BiSeNetV2
5454
Metadata:

‎configs/bisenetv2/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes.py‎

Lines changed: 61 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -3,7 +3,67 @@
33
'../_base_/datasets/cityscapes_1024x1024.py',
44
'../_base_/default_runtime.py', '../_base_/schedules/schedule_160k.py'
55
]
6-
sampler = dict(type='OHEMPixelSampler', thresh=0.7, min_kept=10000)
6+
# sampler = dict(type='OHEMPixelSampler', thresh=0.7, min_kept=10000)
7+
norm_cfg = dict(type='SyncBN', requires_grad=True)
8+
model = dict(
9+
decode_head=dict(
10+
sampler=dict(type='OHEMPixelSampler', thresh=0.7, min_kept=10000)),
11+
auxiliary_head=[
12+
dict(
13+
type='FCNHead',
14+
in_channels=16,
15+
channels=16,
16+
num_convs=2,
17+
num_classes=19,
18+
in_index=1,
19+
norm_cfg=norm_cfg,
20+
concat_input=False,
21+
align_corners=False,
22+
sampler=dict(type='OHEMPixelSampler', thresh=0.7, min_kept=10000),
23+
loss_decode=dict(
24+
type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)),
25+
dict(
26+
type='FCNHead',
27+
in_channels=32,
28+
channels=64,
29+
num_convs=2,
30+
num_classes=19,
31+
in_index=2,
32+
norm_cfg=norm_cfg,
33+
concat_input=False,
34+
align_corners=False,
35+
sampler=dict(type='OHEMPixelSampler', thresh=0.7, min_kept=10000),
36+
loss_decode=dict(
37+
type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)),
38+
dict(
39+
type='FCNHead',
40+
in_channels=64,
41+
channels=256,
42+
num_convs=2,
43+
num_classes=19,
44+
in_index=3,
45+
norm_cfg=norm_cfg,
46+
concat_input=False,
47+
align_corners=False,
48+
sampler=dict(type='OHEMPixelSampler', thresh=0.7, min_kept=10000),
49+
loss_decode=dict(
50+
type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)),
51+
dict(
52+
type='FCNHead',
53+
in_channels=128,
54+
channels=1024,
55+
num_convs=2,
56+
num_classes=19,
57+
in_index=4,
58+
norm_cfg=norm_cfg,
59+
concat_input=False,
60+
align_corners=False,
61+
sampler=dict(type='OHEMPixelSampler', thresh=0.7, min_kept=10000),
62+
loss_decode=dict(
63+
type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)),
64+
],
65+
)
66+
767
lr_config = dict(warmup='linear', warmup_iters=1000)
868
optimizer = dict(lr=0.05)
969
data = dict(

0 commit comments

Comments
 (0)