mirrors
/
EasyCV
mirror of https://github.com/alibaba/EasyCV.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
EasyCV/docs/source/model_zoo_ssl.md

67 lines
8.5 KiB
Markdown
Raw Normal View History

initial commit
2022-04-02 20:01:06 +08:00
# Self-supervised Learning Model Zoo
## Pretrained models
### MAE
Pretrained on **ImageNet** dataset.
| Config | Epochs | Download |
| ------------------------------------------------------------ | ------ | ------------------------------------------------------------ |
feat: readthedocs support & tutorial docs refinement (#29) * readthedocs build support * change opencv version from 4.4.0.40 to 4.1.2.30 * update config file path with github url * fix: clone AutoAugmentPolicy from torchvision error * reorganize requirements file * add model compression tutorial and cls modelzoo to readme.md * add optional req
2022-04-26 10:38:58 +08:00
| [mae_vit_base_patch16_8xb64_400e](https://github.com/alibaba/EasyCV/tree/master/configs/selfsup/mae/mae_vit_base_patch16_8xb64_400e.py) | 400 | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/mae/vit-b-400/pretrain_400.pth) |
| [mae_vit_base_patch16_8xb64_1600e](https://github.com/alibaba/EasyCV/tree/master/configs/selfsup/mae/mae_vit_base_patch16_8xb64_1600e.py) | 1600 | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/mae/vit-b-1600/pretrain_1600.pth) |
| [mae_vit_large_patch16_8xb32_1600e](https://github.com/alibaba/EasyCV/tree/master/configs/selfsup/mae/mae_vit_large_patch16_8xb32_1600e.py) | 1600 | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/mae/vit-l-1600/pretrain_1600.pth) |
initial commit
2022-04-02 20:01:06 +08:00
### DINO
Pretrained on **ImageNet** dataset.
| Config | Epochs | Download |
| ------------------------------------------------------------ | ------ | ------------------------------------------------------------ |
feat: readthedocs support & tutorial docs refinement (#29) * readthedocs build support * change opencv version from 4.4.0.40 to 4.1.2.30 * update config file path with github url * fix: clone AutoAugmentPolicy from torchvision error * reorganize requirements file * add model compression tutorial and cls modelzoo to readme.md * add optional req
2022-04-26 10:38:58 +08:00
| [dino_deit_small_p16_8xb32_100e](https://github.com/alibaba/EasyCV/tree/master/configs/selfsup/dino/dino_deit_small_p16_8xb32_100e_tfrecord.py) | 100 | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/dino_deit_small/epoch_100.pth) - [log](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/dino_deit_small/log.txt) |
initial commit
2022-04-02 20:01:06 +08:00
### MoBY
Pretrained on **ImageNet** dataset.
| Config | Epochs | Download |
| ------------------------------------------------------------ | ------ | ------------------------------------------------------------ |
feat: readthedocs support & tutorial docs refinement (#29) * readthedocs build support * change opencv version from 4.4.0.40 to 4.1.2.30 * update config file path with github url * fix: clone AutoAugmentPolicy from torchvision error * reorganize requirements file * add model compression tutorial and cls modelzoo to readme.md * add optional req
2022-04-26 10:38:58 +08:00
| [moby_deit_small_p16_4xb128_300e](https://github.com/alibaba/EasyCV/tree/master/configs/selfsup/moby/moby_deit_small_p16_4xb128_300e_tfrecord.py) | 300 | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/moby_deit_small_p16/epoch_300.pth) - [log](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/moby_deit_small_p16/log.txt) |
initial commit
2022-04-02 20:01:06 +08:00
### MoCo V2
Pretrained on **ImageNet** dataset.
| Config | Epochs | Download |
| ------------------------------------------------------------ | ------ | ------------------------------------------------------------ |
feat: readthedocs support & tutorial docs refinement (#29) * readthedocs build support * change opencv version from 4.4.0.40 to 4.1.2.30 * update config file path with github url * fix: clone AutoAugmentPolicy from torchvision error * reorganize requirements file * add model compression tutorial and cls modelzoo to readme.md * add optional req
2022-04-26 10:38:58 +08:00
| [mocov2_resnet50_8xb32_200e](https://github.com/alibaba/EasyCV/tree/master/configs/selfsup/mocov2/mocov2_rn50_8xb32_200e_tfrecord.py) | 200 | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/mocov2_r50/epoch_200.pth) - [log](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/mocov2_r50/log.txt) |
initial commit
2022-04-02 20:01:06 +08:00
### SwAV
Pretrained on **ImageNet** dataset.
| Config | Epochs | Download |
| ------------------------------------------------------------ | ------ | ------------------------------------------------------------ |
feat: readthedocs support & tutorial docs refinement (#29) * readthedocs build support * change opencv version from 4.4.0.40 to 4.1.2.30 * update config file path with github url * fix: clone AutoAugmentPolicy from torchvision error * reorganize requirements file * add model compression tutorial and cls modelzoo to readme.md * add optional req
2022-04-26 10:38:58 +08:00
| [swav_resnet50_8xb32_200e](https://github.com/alibaba/EasyCV/tree/master/configs/selfsup/swav/swav_rn50_8xb32_200e_tfrecord.py) | 200 | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/swav_r50/epoch_200.pth) - [log](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/swav_r50/log.txt) |
initial commit
2022-04-02 20:01:06 +08:00
## Benchmarks
remove release_test configs and fix url link od model zoo
2022-04-07 14:28:50 +08:00
For detailed usage of benchmark tools, please refer to benchmark [README.md](../../benchmarks/selfsup/README.md).
initial commit
2022-04-02 20:01:06 +08:00
### ImageNet Linear Evaluation
| Algorithm | Linear Eval Config | Pretrained Config | Top-1 (%) | Download |
| --------- | ------------------------------------------------------------ | ------------------------------------------------------------ | --------- | ------------------------------------------------------------ |
feat: readthedocs support & tutorial docs refinement (#29) * readthedocs build support * change opencv version from 4.4.0.40 to 4.1.2.30 * update config file path with github url * fix: clone AutoAugmentPolicy from torchvision error * reorganize requirements file * add model compression tutorial and cls modelzoo to readme.md * add optional req
2022-04-26 10:38:58 +08:00
| SwAV | [swav_resnet50_8xb2048_20e_feature](../../benchmarks/selfsup/classification/imagenet/swav_r50_8xb2048_20e_feature.py) | [swav_resnet50_8xb32_200e](https://github.com/alibaba/EasyCV/tree/master/configs/selfsup/swav/swav_rn50_8xb32_200e_tfrecord.py) | 73.618 | [log](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/imagenet_linear_eval/swav_r50_linear_eval/20220216_101719.log.json) |
| DINO | [dino_deit_small_p16_8xb2048_20e_feature](../../benchmarks/selfsup/classification/imagenet/dino_deit_small_p16_8xb2048_20e_feature.py) | [dino_deit_small_p16_8xb32_100e](https://github.com/alibaba/EasyCV/tree/master/configs/selfsup/dino/dino_deit_small_p16_8xb32_100e_tfrecord.py) | 71.248 | [log](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/imagenet_linear_eval/dino_deit_small_linear_eval/20220215_141403.log.json) |
| MoBY | [moby_deit_small_p16_8xb2048_30e_feature](../../benchmarks/selfsup/classification/imagenet/moby_deit_small_p16_8xb2048_30e_feature.py) | [moby_deit_small_p16_4xb128_300e](https://github.com/alibaba/EasyCV/tree/master/configs/selfsup/moby/moby_deit_small_p16_4xb128_300e_tfrecord.py) | 72.214 | [log](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/imagenet_linear_eval/moby_deit_small_p16_linear_eval/20220414_134929.log.json) |
| MoCo-v2 | [mocov2_resnet50_8xb2048_40e_feature](../../benchmarks/selfsup/classification/imagenet/mocov2_r50_8xb2048_40e_feature.py) | [mocov2_resnet50_8xb32_200e](https://github.com/alibaba/EasyCV/tree/master/configs/selfsup/mocov2/mocov2_rn50_8xb32_200e_tfrecord.py) | 66.8 | [log](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/imagenet_linear_eval/mocov2_r50_linear_eval/20220214_143738.log.json) |
initial commit
2022-04-02 20:01:06 +08:00
### ImageNet Finetuning
| Algorithm | Fintune Config | Pretrained Config | Top-1 (%) | Download |
| --------- | ------------------------------------------------------------ | ------------------------------------------------------------ | --------- | ------------------------------------------------------------ |
feat: readthedocs support & tutorial docs refinement (#29) * readthedocs build support * change opencv version from 4.4.0.40 to 4.1.2.30 * update config file path with github url * fix: clone AutoAugmentPolicy from torchvision error * reorganize requirements file * add model compression tutorial and cls modelzoo to readme.md * add optional req
2022-04-26 10:38:58 +08:00
| **MAE** | [mae_vit_base_patch16_8xb64_100e_lrdecay075_fintune](../../benchmarks/selfsup/classification/imagenet/mae_vit_base_patch16_8xb64_100e_lrdecay075_fintune.py) | [mae_vit_base_patch16_8xb64_400e](https://github.com/alibaba/EasyCV/tree/master/configs/selfsup/mae/mae_vit_base_patch16_8xb64_400e.py) | 83.13 | [fintune model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/mae/vit-b-400/fintune_400.pth) - [log](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/mae/vit-b-400/20220126_171312.log.json)|
Doc/update mae benchmark (#50) * update mae benchmark, we fintune the model four times, and chose the highest one as the result. we pretrain and fintune the model only by 8 V100 which is a little different from official one.
2022-05-05 20:17:03 +08:00
| | [mae_vit_base_patch16_8xb64_100e_lrdecay065_fintune](../../benchmarks/selfsup/classification/imagenet/mae_vit_base_patch16_8xb64_100e_lrdecay065_fintune.py) | [mae_vit_base_patch16_8xb64_1600e](https://github.com/alibaba/EasyCV/tree/master/configs/selfsup/mae/mae_vit_base_patch16_8xb64_1600e.py) | 83.55 | [fintune model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/mae/vit-b-1600/fintune_1600.pth) - [log](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/mae/vit-b-1600/20220426_101532.log.json)|
| | [mae_vit_large_patch16_8xb16_50e_lrdecay075_fintune](../../benchmarks/selfsup/classification/imagenet/mae_vit_large_patch16_8xb16_50e_lrdecay075_fintune.py) | [mae_vit_large_patch16_8xb32_1600e](https://github.com/alibaba/EasyCV/tree/master/configs/selfsup/mae/mae_vit_large_patch16_8xb32_1600e.py) | 85.70 | [fintune model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/mae/vit-l-1600/fintune_1600.pth) - [log](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/selfsup/mae/vit-l-1600/20220427_150629.log.json)|