add configs for vit backbone plus decode_heads #520

xiexinch · 2021-04-26T06:17:56Z

No description provided.

…into vit_plus_heads

codecov · 2021-04-26T06:58:01Z

Codecov Report

Merging #520 (8db5cdd) into master (98067be) will increase coverage by 0.05%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #520      +/-   ##
==========================================
+ Coverage   85.77%   85.83%   +0.05%     
==========================================
  Files         103      103              
  Lines        5307     5308       +1     
  Branches      857      858       +1     
==========================================
+ Hits         4552     4556       +4     
+ Misses        583      581       -2     
+ Partials      172      171       -1

Flag	Coverage Δ
unittests	`85.81% <100.00%> (+0.04%)`	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
mmseg/models/necks/multilevel_neck.py	`100.00% <100.00%> (ø)`
mmseg/datasets/builder.py	`89.33% <0.00%> (+3.43%)`	⬆️

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 98067be...8db5cdd. Read the comment docs.

xvjiarui · 2021-04-26T16:57:16Z

configs/_base_/models/upernet_vit-d16.py

+        norm_cfg=dict(type='LN'),
+        act_cfg=dict(type='GELU'),
+        norm_eval=False),
+    neck=dict(type='MultiLevelNeck', in_channels=[768], out_channels=768),


We may use [0.5, 1, 2, 4] scale.

xvjiarui · 2021-04-26T21:43:44Z

configs/_base_/models/deeplabv3_vit-d16.py

+    pretrained='https://github.com/rwightman/pytorch-image-models/releases/\
+download/v0.1-vitjx/jx_vit_base_p16_224-80ecf9dd.pth',


We may also add deit-s and deit-b

xvjiarui · 2021-04-27T05:56:28Z

configs/_base_/models/upernet_vit-d16.py

@@ -0,0 +1,54 @@
+# model settings


The config needs to be renamed.

xvjiarui · 2021-04-29T06:33:26Z

configs/_base_/models/deeplabv3_vit-d16.py

+        qk_scale=None,
+        drop_rate=0.0,
+        attn_drop_rate=0.0,
+        norm_cfg=dict(type='LN'),


In #524，eps=1e-6.

xvjiarui · 2021-04-30T07:46:08Z

configs/vit/upernet_deitS-d16-512x512_80k_ade20k.py

+            'absolute_pos_embed': dict(decay_mult=0.),
+            'relative_position_bias_table': dict(decay_mult=0.),


These keys are not in the ViT. They should be pos_embed and cls_token.

xvjiarui · 2021-04-30T07:46:58Z

configs/vit/upernet_deitS-d16-512x512_80k_ade20k.py

+# By default, models are trained on 8 GPUs with 2 images per GPU
+data = dict(samples_per_gpu=2)
+
+find_unused_parameters = True


We may remove these since it will slow down the training and make bugs hard to be found.

…into vit_plus_heads

xvjiarui · 2021-05-06T04:05:27Z

configs/vit/README.md

+
+| Method  | Backbone | Crop Size | Lr schd | Mem (GB) | Inf time (fps) |  mIoU | mIoU(ms+flip) | config                                                                                                                     | download                                                                                                                                                                                                                                                                                                                                               |
+| ------- | -------- | --------- | ------: | -------- | -------------- | ----: | ------------: | -------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
+| UPerNet | Vit     | 512x1024  |   40000 |       |           | |          | |


xvjiarui · 2021-05-06T04:05:35Z

configs/vit/README.md

+| Method  | Backbone | Crop Size | Lr schd | Mem (GB) | Inf time (fps) |  mIoU | mIoU(ms+flip) | config                                                                                                                     | download                                                                                                                                                                                                                                                                                                                                               |
+| ------- | -------- | --------- | ------: | -------- | -------------- | ----: | ------------: | -------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
+| UPerNet | Vit     | 512x1024  |   40000 |       |           | |          | |
+| UPerNet | Deit-S  | 512x1024  |   40000 |       |           | |          | |


xvjiarui · 2021-05-12T02:36:40Z

configs/vit/README.md

+| UPerNet | Vit     | 512x512  |   80000 |       |           | |          | |
+| UPerNet | Deit-S  | 512x512  |   80000 |       |           | |          | |
+| UPerNet | Deit-B  | 512x512  |   80000 |       |           | |          | |


We may also add 160k.

xvjiarui · 2021-05-12T02:38:43Z

configs/_base_/models/fcn_vit-d16.py

+    pretrained='https://github.com/rwightman/pytorch-image-models/releases/\
+download/v0.1-vitjx/jx_vit_base_p16_224-80ecf9dd.pth',


We may use # noqa to bypass the line breaking.

xvjiarui · 2021-05-18T06:07:48Z

configs/_base_/models/deeplabv3_vit-d16.py

@@ -0,0 +1,55 @@
+# model settings


Rename to ***_vit-b16.py

xvjiarui · 2021-05-18T06:08:30Z

configs/vit/deeplabv3_deitB-d16-512x1024_40k_cityscapes.py

@@ -0,0 +1,36 @@
+_base_ = [


Rename to ***_deit-b16.py

xvjiarui · 2021-05-18T06:08:43Z

configs/vit/deeplabv3_deitB-d16-512x512_80k_ade20k.py

+]
+
+model = dict(
+    pretrained='https://dl.fbaipublicfiles.com/deit/\


…tation into vit_plus_heads

xvjiarui · 2021-06-21T03:27:23Z

Are results updated?

xiexinch · 2021-06-21T03:37:10Z

Are results updated?

Not yet, these configs are not correct for the latest vit, I need some time to modify them and test the checkpoints.
I will update soo...

…into vit_plus_heads

Junjun2016 · 2021-06-28T13:05:05Z

configs/_base_/models/upernet_vit-b16.py

@@ -0,0 +1,58 @@
+# model settings


Rename the config.
There is MultiLevelNeck (MLN) in this config.

Junjun2016 · 2021-06-28T13:29:22Z

mmseg/models/necks/multilevel_neck.py

+        for m in self.modules():
+            if isinstance(m, nn.Conv2d):
+                xavier_init(m, distribution='uniform')
+


Why use xavier_init for Conv2d?
kaiming_init for ConvModule is used in MMCV.

Followed FPN

Junjun2016 · 2021-06-28T14:29:33Z

configs/vit/upernet_vit-b16_neck_512x512_80k_ade20k.py

+
+model = dict(
+    decode_head=dict(num_classes=150), auxiliary_head=dict(num_classes=150))
+


drop_path_rate?

ViT doesn't use drop path.

Junjun2016 · 2021-06-28T14:29:49Z

configs/vit/upernet_vit-b16_neck_512x512_160k_ade20k.py

+
+model = dict(
+    decode_head=dict(num_classes=150), auxiliary_head=dict(num_classes=150))
+


drop_path_rate?

ViT doesn't use drop path.

configs/vit/upernet_deit-s16_neck_512x512_160k_ade20k.py

Junjun2016 · 2021-06-28T14:33:27Z

configs/vit/upernet_deit-s16_neck_ln-backbone_512x512_160k_ade20k.py

+
+model = dict(
+    pretrained='https://dl.fbaipublicfiles.com/deit/deit_small_patch16_224-cd65a155.pth',  # noqa
+    backbone=dict(num_heads=6, embed_dims=384, drop_path_rate=0.1, final_norm=True), # noqa


Is the # noqa necessary?

xvjiarui · 2021-06-28T14:43:02Z

configs/vit/upernet_deit-s16_neck_ln-backbone_512x512_160k_ade20k.py

+
+model = dict(
+    pretrained='https://dl.fbaipublicfiles.com/deit/deit_small_patch16_224-cd65a155.pth',  # noqa
+    backbone=dict(num_heads=6, embed_dims=384, drop_path_rate=0.1, final_norm=True), # noqa


Suggested change

backbone=dict(num_heads=6, embed_dims=384, drop_path_rate=0.1, final_norm=True), # noqa

backbone=dict(num_heads=6, embed_dims=384, drop_path_rate=0.1, final_norm=True),

xvjiarui · 2021-06-28T14:56:51Z

configs/vit/README.md

+| ------- | -------- | --------- | ------: | -------- | -------------- | ----: | ------------: | ---------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| UPerNet | ViT-B + neck  | 512x512  | 80000  | 9.20 | 6.94  | 47.71 | 49.51 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/vit/upernet_vit-b16_neck_512x512_80k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_vit-b16_neck_512x512_80k_ade20k-0403cee1.pth) &#124; [log](https://download.openmmlab.com/mmsegmentation/v0.5/vit/20210624_130547.log.json) |
+| UPerNet | ViT-B + neck  | 512x512  | 160000 | 9.20 | 7.58  | 46.75 | 48.46 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/vit/upernet_vit-b16_neck_512x512_160k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_vit-b16_neck_512x512_160k_ade20k-852fa768.pth) &#124; [log](https://download.openmmlab.com/mmsegmentation/v0.5/vit/20210623_192432.log.json) |
+| UPerNet | ViT-B + norm  | 512x512  | 160000 | 9.21 | 6.82  | 47.73 | 49.95 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/vit/upernet_vit-b16_neck_ln-backbone_512x512_160k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_vit-b16_neck_ln-backbone_512x512_160k_ade20k-f444c077.pth) &#124; [log](https://download.openmmlab.com/mmsegmentation/v0.5/vit/20210621_172828.log.json) |


Suggested change

| UPerNet | ViT-B + norm | 512x512 | 160000 | 9.21 | 6.82 | 47.73 | 49.95 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/vit/upernet_vit-b16_neck_ln-backbone_512x512_160k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_vit-b16_neck_ln-backbone_512x512_160k_ade20k-f444c077.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/vit/20210621_172828.log.json) |

| UPerNet | ViT-B + LN +MLN | 512x512 | 160000 | 9.21 | 6.82 | 47.73 | 49.95 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/vit/upernet_vit-b16_ln_mln_512x512_160k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_vit-b16_neck_ln-backbone_512x512_160k_ade20k-f444c077.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/vit/20210621_172828.log.json) |

xvjiarui · 2021-07-01T05:35:34Z

configs/vit/README.md

+
+| Method  | Backbone | Crop Size | Lr schd | Mem (GB) | Inf time (fps) |  mIoU | mIoU(ms+flip) | config                                                                                                                 | download                                                                                                                                                                                                                                                                                                                               |
+| ------- | -------- | --------- | ------: | -------- | -------------- | ----: | ------------: | ---------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| UPerNet | ViT-B + neck  | 512x512  | 80000  | 9.20 | 6.94  | 47.71 | 49.51 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/vit/upernet_vit-b16_neck_512x512_80k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_vit-b16_neck_512x512_80k_ade20k/upernet_vit-b16_neck_512x512_80k_ade20k-0403cee1.pth) &#124; [log](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_vit-b16_neck_512x512_80k_ade20k/20210624_130547.log.json) |


Suggested change

| UPerNet | ViT-B + neck | 512x512 | 80000 | 9.20 | 6.94 | 47.71 | 49.51 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/vit/upernet_vit-b16_neck_512x512_80k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_vit-b16_neck_512x512_80k_ade20k/upernet_vit-b16_neck_512x512_80k_ade20k-0403cee1.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_vit-b16_neck_512x512_80k_ade20k/20210624_130547.log.json) |

| UPerNet | ViT-B + MLN | 512x512 | 80000 | 9.20 | 6.94 | 47.71 | 49.51 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/vit/upernet_vit-b16_neck_512x512_80k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_vit-b16_neck_512x512_80k_ade20k/upernet_vit-b16_neck_512x512_80k_ade20k-0403cee1.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_vit-b16_neck_512x512_80k_ade20k/20210624_130547.log.json) |

xvjiarui · 2021-07-01T05:36:17Z

configs/vit/README.md

+| UPerNet | ViT-B +  LN +MLN  | 512x512  | 160000 | 9.21 | 6.82  | 47.73 | 49.95 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/vit/upernet_vit-b16_ln_mln_512x512_160k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_vit-b16_ln_mln_512x512_160k_ade20k/upernet_vit-b16_ln_mln_512x512_160k_ade20k-f444c077.pth) &#124; [log](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_vit-b16_ln_mln_512x512_160k_ade20k/20210621_172828.log.json) |
+| UPerNet | DeiT-S        | 512x512  | 80000  | 4.68 | 29.85 | 42.96 | 43.79 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/vit/upernet_deit-s16_512x512_80k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_deit-s16_512x512_80k_ade20k/upernet_deit-s16_512x512_80k_ade20k-afc93ec2.pth) &#124; [log](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_deit-s16_512x512_80k_ade20k/20210624_095228.log.json) |
+| UPerNet | DeiT-S        | 512x512  | 160000 | 4.68 | 29.19 | 42.87 | 43.79 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/vit/upernet_deit-s16_512x512_160k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_deit-s16_512x512_160k_ade20k/upernet_deit-s16_512x512_160k_ade20k-5110d916.pth) &#124; [log](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_deit-s16_512x512_160k_ade20k/20210621_160903.log.json) |
+| UPerNet | DeiT-S + neck | 512x512  | 160000 | 5.69 | 11.18 | 43.82 | 45.07 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/vit/upernet_deit-s16_neck_512x512_160k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_deit-s16_neck_512x512_160k_ade20k/upernet_deit-s16_neck_512x512_160k_ade20k-fb9a5dfb.pth) &#124; [log](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_deit-s16_neck_512x512_160k_ade20k/20210621_161021.log.json) |


Suggested change

| UPerNet | DeiT-S + neck | 512x512 | 160000 | 5.69 | 11.18 | 43.82 | 45.07 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/vit/upernet_deit-s16_neck_512x512_160k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_deit-s16_neck_512x512_160k_ade20k/upernet_deit-s16_neck_512x512_160k_ade20k-fb9a5dfb.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_deit-s16_neck_512x512_160k_ade20k/20210621_161021.log.json) |

| UPerNet | DeiT-S + MLN | 512x512 | 160000 | 5.69 | 11.18 | 43.82 | 45.07 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/vit/upernet_deit-s16_neck_512x512_160k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_deit-s16_neck_512x512_160k_ade20k/upernet_deit-s16_neck_512x512_160k_ade20k-fb9a5dfb.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/vit/upernet_deit-s16_neck_512x512_160k_ade20k/20210621_161021.log.json) |

xvjiarui · 2021-07-01T05:36:39Z

configs/vit/upernet_deit-b16_512x512_160k_ade20k.py

@@ -0,0 +1,6 @@
+_base_ = './upernet_vit-b16_neck_512x512_160k_ade20k.py'


Suggested change

_base_ = './upernet_vit-b16_neck_512x512_160k_ade20k.py'

_base_ = './upernet_vit-b16_mln_512x512_160k_ade20k.py'

* add config * add cityscapes config * add default value to docstring * fix lint * add deit-s and deit-b * add readme * add eps at norm_cfg * add drop_path_rate experiment * add deit case at init_weight * add upernet result * update result and add upernet 160k config * update upernet result and fix settings * Update iters number * update result and delete some configs * fix import error * fix drop_path_rate * update result and restore config * update benchmark result * remove cityscapes exp * remove neck * neck exp * add more configs * fix init error * fix ffn setting * update result * update results * update result * update results and fill table * delete or rename configs * fix link delimiter * rename configs and fix link * rename neck to mln

fix typo

xiexinch added 4 commits April 26, 2021 12:05

add config

1c66609

add cityscapes config

a439a40

add default value to docstring

b7aa7ee

Merge branch 'master' of https://github.com/open-mmlab/mmsegmentation …

4e8383c

…into vit_plus_heads

fix lint

e7d6243

xvjiarui reviewed Apr 26, 2021

View reviewed changes

add deit-s and deit-b

0ce8958

xvjiarui reviewed Apr 27, 2021

View reviewed changes

configs/_base_/models/upernet_vit-d16.py Outdated

@@ -0,0 +1,54 @@

# model settings

Copy link

Collaborator

xvjiarui Apr 27, 2021

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The config needs to be renamed.

add readme

6b40465

xvjiarui reviewed Apr 29, 2021

View reviewed changes

add eps at norm_cfg

6d0ab21

xvjiarui reviewed Apr 30, 2021

View reviewed changes

xiexinch added 4 commits May 6, 2021 10:27

Merge branch 'master' of https://github.com/open-mmlab/mmsegmentation …

c9c6596

…into vit_plus_heads

add drop_path_rate experiment

f7b8c18

add deit case at init_weight

d605af6

add upernet result

425cac7

xvjiarui reviewed May 12, 2021

View reviewed changes

xiexinch added 3 commits May 12, 2021 12:53

update result and add upernet 160k config

dd6856e

update upernet result and fix settings

bd86b64

Update iters number

3f8db05

xvjiarui reviewed May 18, 2021

View reviewed changes

clownrat6 mentioned this pull request May 18, 2021

How can I use ViT? #525

Closed

xiexinch added 4 commits May 18, 2021 15:20

update result and delete some configs

4d77761

fix import error

656c167

fix drop_path_rate

7d13836

update result and restore config

8931d7f

Merge branch 'vit_plus_heads' of https://github.com/xiexinch/mmsegmen…

e7a6637

…tation into vit_plus_heads

xiexinch added 9 commits June 21, 2021 15:03

fix init error

319f56e

fix ffn setting

c4aa7b6

update result

8457e67

update results

b0f9e29

update result

74424aa

update results and fill table

8c11f62

Merge branch 'master' of https://github.com/open-mmlab/mmsegmentation …

637356d

…into vit_plus_heads

delete or rename configs

31ba78d

fix link delimiter

53ae35c

Junjun2016 self-requested a review June 28, 2021 09:42

Junjun2016 reviewed Jun 28, 2021

View reviewed changes

configs/vit/upernet_deit-s16_neck_512x512_160k_ade20k.py Outdated Show resolved Hide resolved

Junjun2016 reviewed Jun 28, 2021

View reviewed changes

xvjiarui reviewed Jun 28, 2021

View reviewed changes

rename configs and fix link

451ae11

xvjiarui reviewed Jul 1, 2021

View reviewed changes

rename neck to mln

8db5cdd

Junjun2016 approved these changes Jul 1, 2021

View reviewed changes

Junjun2016 merged commit 737544f into open-mmlab:master Jul 1, 2021

sibozhang pushed a commit to sibozhang/mmsegmentation that referenced this pull request Mar 22, 2024

Update README.md (open-mmlab#520)

b769492

fix typo

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

add configs for vit backbone plus decode_heads #520

add configs for vit backbone plus decode_heads #520

xiexinch commented Apr 26, 2021

codecov bot commented Apr 26, 2021 •

edited

Loading

xvjiarui Apr 26, 2021

xvjiarui Apr 26, 2021

xvjiarui Apr 27, 2021

xvjiarui Apr 29, 2021

xvjiarui Apr 30, 2021

xvjiarui Apr 30, 2021

xvjiarui May 6, 2021

xvjiarui May 6, 2021

xvjiarui May 12, 2021

xvjiarui May 12, 2021

xvjiarui May 18, 2021

xvjiarui May 18, 2021

xvjiarui May 18, 2021

xvjiarui commented Jun 21, 2021

xiexinch commented Jun 21, 2021

Junjun2016 Jun 28, 2021

Junjun2016 Jun 28, 2021

xvjiarui Jun 28, 2021

Junjun2016 Jun 28, 2021

xvjiarui Jun 28, 2021

Junjun2016 Jun 28, 2021

xvjiarui Jun 28, 2021

Junjun2016 Jun 28, 2021

xvjiarui Jun 28, 2021

xvjiarui Jun 28, 2021

xvjiarui Jul 1, 2021

xvjiarui Jul 1, 2021

xvjiarui Jul 1, 2021

		pretrained='https://github.com/rwightman/pytorch-image-models/releases/\
		download/v0.1-vitjx/jx_vit_base_p16_224-80ecf9dd.pth',

		'absolute_pos_embed': dict(decay_mult=0.),
		'relative_position_bias_table': dict(decay_mult=0.),


		model = dict(
		decode_head=dict(num_classes=150), auxiliary_head=dict(num_classes=150))

	backbone=dict(num_heads=6, embed_dims=384, drop_path_rate=0.1, final_norm=True), # noqa
	backbone=dict(num_heads=6, embed_dims=384, drop_path_rate=0.1, final_norm=True),

		@@ -0,0 +1,6 @@
		_base_ = './upernet_vit-b16_neck_512x512_160k_ade20k.py'

	_base_ = './upernet_vit-b16_neck_512x512_160k_ade20k.py'
	_base_ = './upernet_vit-b16_mln_512x512_160k_ade20k.py'

add configs for vit backbone plus decode_heads #520

add configs for vit backbone plus decode_heads #520

Conversation

xiexinch commented Apr 26, 2021

codecov bot commented Apr 26, 2021 • edited Loading

Codecov Report

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

xvjiarui commented Jun 21, 2021

xiexinch commented Jun 21, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

codecov bot commented Apr 26, 2021 •

edited

Loading