mirrors
/
mmsegmentation
mirror of https://github.com/open-mmlab/mmsegmentation.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

[Fix] update pp_mobileseg ckpt links (#3254)

pull/3261/head
谢昕辰 2023-08-10 18:56:30 +08:00 committed by GitHub
parent e458a467d6
commit 9801c62fd7
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
3 changed files with 28 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
projects/pp_mobileseg/README.md
View File

@ -30,10 +30,18 @@ Extensive experiments show that PP-MobileSeg achieves a superior params-accuracy
### ADE20K
| Model | Backbone | Training Iters | Batchsize | Train Resolution | mIoU(%) | latency(ms)\* | params(M) | config | Links |
| ----------------- | ----------------- | -------------- | --------- | ---------------- | ------- | ------------- | --------- | ------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| PP-MobileSeg-Base | StrideFormer-Base | 80000 | 32 | 512x512 | 41.57% | 265.5 | 5.62 | [config](https://github.com/Yang-Changhui/mmsegmentation/tree/add_ppmobileseg/projects/pp_mobileseg/configs/pp_mobileseg) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/pp_mobileseg/pp_mobileseg_mobilenetv3_2x16_80k_ade20k_512x512_base-ed0be681.pth)\|[log](https://bj.bcebos.com/paddleseg/dygraph/ade20k/pp_mobileseg_base/train.log) |
| PP-MobileSeg-Tiny | StrideFormer-Tiny | 80000 | 32 | 512x512 | 36.39% | 215.3 | 1.61 | [config](https://github.com/Yang-Changhui/mmsegmentation/tree/add_ppmobileseg/projects/pp_mobileseg/configs/pp_mobileseg) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/pp_mobileseg/pp_mobileseg_mobilenetv3_2x16_80k_ade20k_512x512_tiny-e4b35e96.pth)\|[log](https://bj.bcebos.com/paddleseg/dygraph/ade20k/pp_mobileseg_tiny/train.log) |
| Model | Backbone | Training Iters | Batchsize | Train Resolution | mIoU(%) | latency(ms)\* | params(M) | config | Links |
| ----------------- | ----------------- | -------------- | --------- | ---------------- | ------- | ------------- | --------- | ------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| PP-MobileSeg-Base | StrideFormer-Base | 80000 | 32 | 512x512 | 41.57% | 265.5 | 5.62 | [config](https://github.com/Yang-Changhui/mmsegmentation/tree/add_ppmobileseg/projects/pp_mobileseg/configs/pp_mobileseg) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/pp_mobileseg/pp_mobileseg_mobilenetv3_2xb16_3rdparty-base_512x512-ade20k-f12b44f3.pth)\|[log](https://bj.bcebos.com/paddleseg/dygraph/ade20k/pp_mobileseg_base/train.log) |
| PP-MobileSeg-Tiny | StrideFormer-Tiny | 80000 | 32 | 512x512 | 36.39% | 215.3 | 1.61 | [config](https://github.com/Yang-Changhui/mmsegmentation/tree/add_ppmobileseg/projects/pp_mobileseg/configs/pp_mobileseg) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/pp_mobileseg/pp_mobileseg_mobilenetv3_2xb16_3rdparty-tiny_512x512-ade20k-a351ebf5.pth)\|[log](https://bj.bcebos.com/paddleseg/dygraph/ade20k/pp_mobileseg_tiny/train.log) |
## Usage
Same as other models in MMsegmentation, you can run the following command to test the model at ${MMSEG_ROOT}:
```shell
./tools/dist_test.sh projects/pp_mobileseg/configs/pp_mobileseg/pp_mobileseg_mobilenetv3_2x16_80k_ade20k_512x512_base.py checkpoints/pp_mobileseg_mobilenetv3_2xb16_3rdparty-base_512x512-ade20k-f12b44f3.pth 8
```
## Citation

13
projects/pp_mobileseg/configs/pp_mobileseg/pp_mobileseg_mobilenetv3_2x16_80k_ade20k_512x512_base.py
View File

@ -2,12 +2,17 @@ _base_ = [
'../_base_/models/pp_mobile.py', '../_base_/datasets/ade20k.py',
'../_base_/default_runtime.py', '../_base_/schedules/schedule_80k.py'
]
checkpoint = './models/pp_mobile_base.pth'
# the custom import path is determined by your workspace path (i.e., where you run the command from) # noqa
custom_imports = dict(
imports=[
'projects.pp_mobileseg.backbones', 'projects.pp_mobileseg.decode_head'
],
allow_failed_imports=False)
checkpoint = 'https://download.openmmlab.com/mmsegmentation/v0.5/pp_mobileseg/pp_mobileseg_mobilenetv3_3rdparty-base-ed0be681.pth' # noqa
crop_size = (512, 512)
data_preprocessor = dict(size=crop_size, test_cfg=dict(size_divisor=32))
norm_cfg = dict(type='SyncBN', requires_grad=True)
model = dict(
data_preprocessor=data_preprocessor,
backbone=dict(init_cfg=dict(type='Pretrained', checkpoint=checkpoint), ),
decode_head=dict(num_classes=150, upsample='intepolate'),
)
backbone=dict(init_cfg=dict(type='Pretrained', checkpoint=checkpoint)),
decode_head=dict(num_classes=150, upsample='intepolate'))

8
projects/pp_mobileseg/configs/pp_mobileseg/pp_mobileseg_mobilenetv3_2x16_80k_ade20k_512x512_tiny.py
View File

@ -2,7 +2,13 @@ _base_ = [
'../_base_/models/pp_mobile.py', '../_base_/datasets/ade20k.py',
'../_base_/default_runtime.py', '../_base_/schedules/schedule_80k.py'
]
checkpoint = './models/pp_mobile_tiny.pth'
# the custom import path is determined by your workspace path (i.e., where you run the command from) # noqa
custom_imports = dict(
imports=[
'projects.pp_mobileseg.backbones', 'projects.pp_mobileseg.decode_head'
],
allow_failed_imports=False)
checkpoint = 'https://download.openmmlab.com/mmsegmentation/v0.5/pp_mobileseg/pp_mobileseg_mobilenetv3_3rdparty-tiny-e4b35e96.pth' # noqa
crop_size = (512, 512)
data_preprocessor = dict(size=crop_size, test_cfg=dict(size_divisor=32))
norm_cfg = dict(type='SyncBN', requires_grad=True)