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

[Enhance] enhance MIM function (#152) 

* [Docs] update mminstall.txt

* [Docs] add MANIFEST.in

* [Docs] add metafile

* [WIP]: temporal commit

* [Enhance] Try to create a symbolic link on windows.

* [Docs] add metefile of algorithms
pull/153/head
Yixiao Fang 2021-12-17 16:39:03 +08:00 committed by GitHub
parent e882fd9c01
commit 9ef4b9a3d8
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
15 changed files with 359 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
MANIFEST.in 100644
View File

@ -0,0 +1,5 @@
include requirements/*.txt
include mmselfsup/VERSION
include mmselfsup/.mim/model-index.yml
recursive-include mmselfsup/.mim/configs *.py *.yml
recursive-include mmselfsup/.mim/tools *.sh *.py

28
configs/selfsup/byol/metafile.yml 100644
View File

@ -0,0 +1,28 @@
Collections:
- Name: BYOL
Metadata:
Training Data: ImageNet-1k
Training Techniques:
- LARS
Training Resources: 8x V100 GPUs
Architecture:
- ResNet
- BYOL
Paper:
URL: https://arxiv.org/abs/2006.07733
Title: "Bootstrap your own latent: A new approach to self-supervised Learning"
README: configs/selfsup/byol/README.md
Models:
- Name: byol_resnet50_8xb32-accum16-coslr-200e_in1k
In Collection: BYOL
Metadata:
Epochs: 200
Batch Size: 256
Results:
- Task: Self-Supervised Image Classification
Dataset: ImageNet-1k
Metrics:
Top 1 Accuracy: 67.68
Config: configs/selfsup/byol/byol_resnet50_8xb32-accum16-coslr-200e_in1k.py
Weights: https://download.openmmlab.com/mmselfsup/byol/byol_resnet50_8xb32-accum16-coslr-200e_in1k_20211213-30dbaef1.pth

29
configs/selfsup/deepcluster/metafile.yml 100644
View File

@ -0,0 +1,29 @@
Collections:
- Name: DeepCluster
Metadata:
Training Data: ImageNet-1k
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Resources: 8x V100 GPUs
Architecture:
- ResNet
- DeepCluster
Paper:
URL: https://arxiv.org/abs/1807.05520
Title: "Deep clustering for unsupervised learning of visual features"
README: configs/selfsup/deepcluster/README.md
Models:
- Name: deepcluster-sobel_resnet50_8xb64-steplr-200e_in1k
In Collection: DeepCluster
Metadata:
Epochs: 200
Batch Size: 512
Results:
- Task: Self-Supervised Image Classification
Dataset: ImageNet-1k
Metrics:
Top 1 Accuracy: 46.92
Config: configs/selfsup/deepcluster/deepcluster-sobel_resnet50_8xb64-steplr-200e_in1k.py
Weights: https://download.openmmlab.com/mmselfsup/deepcluster/deepcluster-sobel_resnet50_8xb64-steplr-200e_in1k-bb8681e2.pth

28
configs/selfsup/densecl/metafile.yml 100644
View File

@ -0,0 +1,28 @@
Collections:
- Name: DenseCL
Metadata:
Training Data: ImageNet-1k
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Resources: 8x V100 GPUs
Architecture:
- ResNet
Paper:
URL: https://arxiv.org/abs/2011.09157
Title: "Dense contrastive learning for self-supervised visual pre-training"
README: configs/selfsup/densecl/README.md
Models:
- Name: densecl_resnet50_8xb32-coslr-200e_in1k
In Collection: DenseCL
Metadata:
Epochs: 200
Batch Size: 256
Results:
- Task: Self-Supervised Image Classification
Dataset: ImageNet-1k
Metrics:
Top 1 Accuracy: 63.34
Config: configs/selfsup/densecl/densecl_resnet50_8xb32-coslr-200e_in1k.py
Weights: https://download.openmmlab.com/mmselfsup/densecl/densecl_resnet50_8xb32-coslr-200e_in1k_20211214-1efb342c.pth

32
configs/selfsup/moco/metafile.yml 100644
View File

@ -0,0 +1,32 @@
Collections:
- Name: MoCo
Metadata:
Training Data: ImageNet-1k
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Resources: 8x V100 GPUs
Architecture:
- ResNet
- MoCo
Paper:
URL: https://arxiv.org/abs/1911.05722
Title: "Momentum Contrast for Unsupervised Visual Representation Learning"
README: configs/selfsup/moco/README.md
Models:
- Name: mocov2_resnet50_8xb32-coslr-200e_in1k
In Collection: MoCo
Metadata:
Epochs: 200
Batch Size: 256
Results:
- Task: Self-Supervised Image Classification
Dataset: ImageNet-1k
Metrics:
Top 1 Accuracy: 67.56
Config: configs/selfsup/moco/mocov2_resnet50_8xb32-coslr-200e_in1k.py
Weights: https://download.openmmlab.com/mmselfsup/moco/mocov2_resnet50_8xb32-coslr-200e_in1k_20211213-7ce8f840.pth
Paper:
URL: https://arxiv.org/abs/2003.04297
Title: "Improved Baselines with Momentum Contrastive Learning"

29
configs/selfsup/npid/metafile.yml 100644
View File

@ -0,0 +1,29 @@
Collections:
- Name: NPID
Metadata:
Training Data: ImageNet-1k
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Resources: 8x V100 GPUs
Architecture:
- ResNet
- NPID
Paper:
URL: https://arxiv.org/abs/1805.01978
Title: "Unsupervised Feature Learning via Non-Parametric Instance Discrimination"
README: configs/selfsup/npid/README.md
Models:
- Name: npid_resnet50_8xb32-steplr-200e_in1k
In Collection: NPID
Metadata:
Epochs: 200
Batch Size: 256
Results:
- Task: Self-Supervised Image Classification
Dataset: ImageNet-1k
Metrics:
Top 1 Accuracy: 58.16
Config: configs/selfsup/npid/npid_resnet50_8xb32-steplr-200e_in1k.py
Weights: https://download.openmmlab.com/mmselfsup/npid/npid_resnet50_8xb32-steplr-200e_in1k_20211213-b5fec6df.pth

28
configs/selfsup/odc/metafile.yml 100644
View File

@ -0,0 +1,28 @@
Collections:
- Name: ODC
Metadata:
Training Data: ImageNet-1k
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Resources: 8x V100 GPUs
Architecture:
- ResNet
Paper:
URL: https://arxiv.org/abs/2006.10645
Title: "Online deep clustering for unsupervised representation learning"
README: configs/selfsup/odc/README.md
Models:
- Name: odc_resnet50_8xb64-steplr-440e_in1k
In Collection: ODC
Metadata:
Epochs: 200
Batch Size: 512
Results:
- Task: Self-Supervised Image Classification
Dataset: ImageNet-1k
Metrics:
Top 1 Accuracy: 53.42
Config: configs/selfsup/odc/odc_resnet50_8xb64-steplr-440e_in1k.py
Weights: https://download.openmmlab.com/mmselfsup/odc/odc_resnet50_8xb64-steplr-440e_in1k-5af5dd0c.pth

28
configs/selfsup/relative_loc/metafile.yml 100644
View File

@ -0,0 +1,28 @@
Collections:
- Name: RelativeLocation
Metadata:
Training Data: ImageNet-1k
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Resources: 8x V100 GPUs
Architecture:
- ResNet
Paper:
URL: https://arxiv.org/abs/1505.05192
Title: "Unsupervised visual representation learning by context prediction"
README: configs/selfsup/relative_loc/README.md
Models:
- Name: relative-loc_resnet50_8xb64-steplr-70e_in1k
In Collection: RelativeLocation
Metadata:
Epochs: 70
Batch Size: 512
Results:
- Task: Self-Supervised Image Classification
Dataset: ImageNet-1k
Metrics:
Top 1 Accuracy: 39.65
Config: configs/selfsup/relative_loc/relative-loc_resnet50_8xb64-steplr-70e_in1k.py
Weights: https://download.openmmlab.com/mmselfsup/relative_loc/relative-loc_resnet50_8xb64-steplr-70e_in1k_20211213-cdd3162f.pth

28
configs/selfsup/rotation_pred/metafile.yml 100644
View File

@ -0,0 +1,28 @@
Collections:
- Name: RotationPrediction
Metadata:
Training Data: ImageNet-1k
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Resources: 8x V100 GPUs
Architecture:
- ResNet
Paper:
URL: https://arxiv.org/abs/1803.07728
Title: "Unsupervised representation learning by predicting image rotations"
README: configs/selfsup/rotation_pred/README.md
Models:
- Name: rotation-pred_resnet50_8xb16-steplr-70e_in1k
In Collection: RotationPrediction
Metadata:
Epochs: 70
Batch Size: 128
Results:
- Task: Self-Supervised Image Classification
Dataset: ImageNet-1k
Metrics:
Top 1 Accuracy: 39.65
Config: configs/selfsup/rotation_pred/rotation-pred_resnet50_8xb16-steplr-70e_in1k.py
Weights: https://download.openmmlab.com/mmselfsup/rotation_pred/rotation-pred_resnet50_8xb16-steplr-70e_in1k_20211213-513972ac.pth

28
configs/selfsup/simclr/metafile.yml 100644
View File

@ -0,0 +1,28 @@
Collections:
- Name: SimCLR
Metadata:
Training Data: ImageNet-1k
Training Techniques:
- LARS
Training Resources: 8x V100 GPUs
Architecture:
- ResNet
- SimCLR
Paper:
URL: https://arxiv.org/abs/2002.05709
Title: "A simple framework for contrastive learning of visual representations"
README: configs/selfsup/simclr/README.md
Models:
- Name: simclr_resnet50_8xb32-coslr-200e_in1k
In Collection: SimCLR
Metadata:
Epochs: 200
Batch Size: 256
Results:
- Task: Self-Supervised Image Classification
Dataset: ImageNet-1k
Metrics:
Top 1 Accuracy: 58.92
Config: configs/selfsup/simclr/simclr_resnet50_8xb32-coslr-200e_in1k.py
Weights: https://download.openmmlab.com/mmselfsup/simclr/simclr_resnet50_8xb32-coslr-200e_in1k_20211213-d0e53669.pth

40
configs/selfsup/simsiam/metafile.yml 100644
View File

@ -0,0 +1,40 @@
Collections:
- Name: SimSiam
Metadata:
Training Data: ImageNet-1k
Training Techniques:
- SGD with Momentum
- Weight Decay
Training Resources: 8x V100 GPUs
Architecture:
- ResNet
Paper:
URL: https://arxiv.org/abs/2011.10566
Title: "Exploring simple siamese representation learning"
README: configs/selfsup/simsiam/README.md
Models:
- Name: simsiam_resnet50_8xb32-coslr-100e_in1k
In Collection: SimSiam
Metadata:
Epochs: 100
Batch Size: 256
Results:
- Task: Self-Supervised Image Classification
Dataset: ImageNet-1k
Metrics:
Top 1 Accuracy: 67.88
Config: configs/selfsup/simsiam/simsiam_resnet50_8xb32-coslr-100e_in1k.py
Weights: https://download.openmmlab.com/mmselfsup/simsiam/simsiam_resnet50_8xb32-coslr-100e_in1k_20211213-925d628c.pth
- Name: simsiam_resnet50_8xb32-coslr-200e_in1k
In Collection: SimSiam
Metadata:
Epochs: 200
Batch Size: 256
Results:
- Task: Self-Supervised Image Classification
Dataset: ImageNet-1k
Metrics:
Top 1 Accuracy: 69.80
Config: configs/selfsup/simsiam/simsiam_resnet50_8xb32-coslr-200e_in1k.py
Weights: https://download.openmmlab.com/mmselfsup/simsiam/simsiam_resnet50_8xb32-coslr-200e_in1k_20211213-b605f9f1.pth

28
configs/selfsup/swav/metafile.yml 100644
View File

@ -0,0 +1,28 @@
Collections:
- Name: SwAV
Metadata:
Training Data: ImageNet-1k
Training Techniques:
- LARS
Training Resources: 8x V100 GPUs
Architecture:
- ResNet
- SwAV
Paper:
URL: https://arxiv.org/abs/2006.09882
Title: "Unsupervised Learning of Visual Features by Contrasting Cluster Assignments"
README: configs/selfsup/swav/README.md
Models:
- Name: swav_resnet50_8xb32-mcrop-2-6-coslr-200e_in1k-224-96
In Collection: SwAV
Metadata:
Epochs: 200
Batch Size: 256
Results:
- Task: Self-Supervised Image Classification
Dataset: ImageNet-1k
Metrics:
Top 1 Accuracy: 70.55
Config: configs/selfsup/swav/swav_resnet50_8xb32-mcrop-2-6-coslr-200e_in1k-224-96.py
Weights: https://download.openmmlab.com/mmselfsup/swav/swav_resnet50_8xb32-mcrop-2-6-coslr-200e_in1k-224-96_20211213-0028900c.pth

12
model-index.yml 100644
View File

@ -0,0 +1,12 @@
Import:
- configs/selfsup/byol/metafile.yml
- configs/selfsup/deepcluster/metafile.yml
- configs/selfsup/densecl/metafile.yml
- configs/selfsup/moco/metafile.yml
- configs/selfsup/npid/metafile.yml
- configs/selfsup/odc/metafile.yml
- configs/selfsup/relative_loc/metafile.yml
- configs/selfsup/rotation_pred/metafile.yml
- configs/selfsup/simclr/metafile.yml
- configs/selfsup/simsiam/metafile.yml
- configs/selfsup/swav/metafile.yml

2
requirements/mminstall.txt
View File

@ -1 +1,3 @@
mmcv-full>=1.3.16
mmdet >= 2.16.0
mmsegmentation >= 0.20.2

16
setup.py
View File

@ -133,8 +133,20 @@ def add_mim_extension():
if mode == 'symlink':
src_relpath = osp.relpath(src_path, osp.dirname(tar_path))
os.symlink(src_relpath, tar_path)
elif mode == 'copy':
try:
os.symlink(src_relpath, tar_path)
except OSError:
# Creating a symbolic link on windows may raise an
# `OSError: [WinError 1314]` due to privilege. If
# the error happens, the src file will be copied
mode = 'copy'
warnings.warn(
f'Failed to create a symbolic link for {src_relpath}, '
f'and it will be copied to {tar_path}')
else:
continue
if mode == 'copy':
if osp.isfile(src_path):
shutil.copyfile(src_path, tar_path)
elif osp.isdir(src_path):