update modelzoo
xieenze
parent
6c7292a495
commit
ad856e041b
|
|
@ -9,24 +9,25 @@ Since we don't have sufficient GPUs to run these large-scale experiments, your c
|
|||
* Experiments with the same batch size are directly comparable in speed.
|
||||
* The table records the implementors who implemented the methods (either by themselves or refactoring from other repos), and the experimenters who performed experiments and reproduced the results. The experimenters should be responsible for the evaluation results on all the benchmarks, and the implementors should be responsible for the implementation as well as the results; If the experimenter is not indicated, an implementator is the experimenter by default.
|
||||
|
||||
<table><thead><tr><th>Method (Implementator)</th><th>Config (Experimenter)</th><th>Remarks</th><th>Download link</th><th>Batch size</th><th>Epochs</th><th>Time per epoch</th></tr></thead><tbody>
|
||||
<tr><td><a href="https://github.com/pytorch/vision/blob/master/torchvision/models/resnet.py" target="_blank" rel="noopener noreferrer">ImageNet</a></td><td>-</td><td>torchvision</td><td><a href="https://drive.google.com/file/d/11xA3TOcbD0qOrwpBfYonEDeseE1wMfBh/view?usp=sharing" target="_blank" rel="noopener noreferrer">imagenet_r50-21352794.pth</a></td><td>-</td><td>-</td><td>-</td></tr>
|
||||
<tr><td>Random</td><td>-</td><td>kaiming</td><td><a href="https://drive.google.com/file/d/1UaFTjd6sbKkZEE-f58Zv30bnx7C1qJBb/view?usp=sharing" target="_blank" rel="noopener noreferrer">random_r50-5d0fa71b.pth</a></td><td>-</td><td>-</td><td>-</td></tr>
|
||||
<tr><td><a href="https://www.cv-foundation.org/openaccess/content_iccv_2015/papers/Doersch_Unsupervised_Visual_Representation_ICCV_2015_paper.pdf" target="_blank" rel="noopener noreferrer">Relative-Loc</a> (<a href="https://github.com/Jiahao000">@Jiahao000</a>)</td><td>selfsup/relative_loc/r50.py</td><td>default</td><td><a href="https://drive.google.com/file/d/1ibk1BI3PFQxZqcxuDfHs3n7JnWKCgl8x/view?usp=sharing" target="_blank" rel="noopener noreferrer">relative_loc_r50-342c9097.pth</a></td><td>512</td><td>70</td><td>21min17s</td></tr>
|
||||
<tr><td><a href="https://arxiv.org/abs/1803.07728" target="_blank" rel="noopener noreferrer">Rotation-Pred</a> (<a href="https://github.com/XiaohangZhan">@XiaohangZhan</a>)</td><td>selfsup/rotation_pred/r50.py</td><td>default</td><td><a href="https://drive.google.com/file/d/1t3oClmIvQ0p8RZ0V5yvQFltzjqBO823Y/view?usp=sharing" target="_blank" rel="noopener noreferrer">rotation_r50-cfab8ebb.pth</a></td><td>128</td><td>70</td><td>49min58s</td></tr>
|
||||
<tr><td><a href="https://arxiv.org/abs/1807.05520" target="_blank" rel="noopener noreferrer">DeepCluster</a> (<a href="https://github.com/XiaohangZhan">@XiaohangZhan</a>)</td><td>selfsup/deepcluster/r50.py</td><td>default</td><td><a href="https://drive.google.com/file/d/1GxgP7pI18JtFxDIC0hnHOanvUYajoLlg/view?usp=sharing" target="_blank" rel="noopener noreferrer">deepcluster_r50-bb8681e2.pth</a></td><td>512</td><td>200</td><td>41min57s</td></tr>
|
||||
<tr><td><a href="https://arxiv.org/abs/1805.01978" target="_blank" rel="noopener noreferrer">NPID</a> (<a href="https://github.com/XiaohangZhan">@XiaohangZhan</a>)</td><td>selfsup/npid/r50.py</td><td>default</td><td><a href="https://drive.google.com/file/d/1sm6I3Y5XnCWdbmeLSF4YupUtPe5nRQMI/view?usp=sharing" target="_blank" rel="noopener noreferrer">npid_r50-dec3df0c.pth</a></td><td>256</td><td>200</td><td>20min5s</td></tr>
|
||||
<tr><td></td><td>selfsup/npid/r50_ensure_neg.py</td><td>ensure_neg=True</td><td><a href="https://drive.google.com/file/d/1FldDrb6kzF3CZ7737mwCXVI6HE2aCSaF/view?usp=sharing" target="_blank" rel="noopener noreferrer">npid_r50_ensure_neg-ce09b7ae.pth</a></td><td></td><td></td><td></td></tr>
|
||||
<tr><td><a href="http://openaccess.thecvf.com/content_CVPR_2020/papers/Zhan_Online_Deep_Clustering_for_Unsupervised_Representation_Learning_CVPR_2020_paper.pdf" target="_blank" rel="noopener noreferrer">ODC</a> (<a href="https://github.com/XiaohangZhan">@XiaohangZhan</a>)</td><td>selfsup/odc/r50_v1.py (<a href="https://github.com/Jiahao000">@Jiahao000</a>)</td><td>default</td><td><a href="https://drive.google.com/file/d/1EdhJeZAyMsD_pEW7uMhLzos5xZLdariN/view?usp=sharing" target="_blank" rel="noopener noreferrer">odc_r50_v1-5af5dd0c.pth</a></td><td>512</td><td>440</td><td>28min22s</td></tr>
|
||||
<tr><td><a href="https://arxiv.org/abs/1911.05722" target="_blank" rel="noopener noreferrer">MoCo</a> (<a href="https://github.com/XiaohangZhan">@XiaohangZhan</a>)</td><td>selfsup/moco/r50_v1.py</td><td>default</td><td><a href="https://drive.google.com/file/d/1ANXfnoT8yBQQBBqR_kQLQorK20l65KMy/view?usp=sharing" target="_blank" rel="noopener noreferrer">moco_r50_v1-4ad89b5c.pth</a></td><td>256</td><td>200</td><td>22min58s</td></tr>
|
||||
<tr><td><a href="https://arxiv.org/abs/2003.04297" target="_blank" rel="noopener noreferrer">MoCo v2</a> (<a href="https://github.com/XiaohangZhan">@XiaohangZhan</a>)</td><td>selfsup/moco/r50_v2.py</td><td>default</td><td><a href="https://drive.google.com/file/d/1ImO8A3uWbrTx21D1IqBDMUQvpN6wmv0d/view?usp=sharing" target="_blank" rel="noopener noreferrer">moco_r50_v2-e3b0c442.pth</a></td><td>256</td><td>200</td><td>25min7s</td></tr>
|
||||
<tr><td><a href="https://arxiv.org/abs/2002.05709" target="_blank" rel="noopener noreferrer">SimCLR</a> (<a href="https://github.com/XiaohangZhan">@XiaohangZhan</a>)</td><td>selfsup/simclr/r50_bs256_ep200.py</td><td>default</td><td><a href="https://drive.google.com/file/d/1aZ43nSdivdNxHbM9DKVoZYVhZ8TNnmPp/view?usp=sharing" target="_blank" rel="noopener noreferrer">simclr_r50_bs256_ep200-4577e9a6.pth</a></td><td>256</td><td>200</td><td>32min13s</td></tr>
|
||||
<tr><td></td><td>selfsup/simclr/r50_bs256_ep200_mocov2_neck.py</td><td>-> MoCo v2 neck</td><td><a href="https://drive.google.com/file/d/1AXpSKqgWfnj6jCgN65BXSTCKFfuIVELa/view?usp=sharing" target="_blank" rel="noopener noreferrer">simclr_r50_bs256_ep200_mocov2_neck-0d6e5ff2.pth</a></td><td></td><td></td><td></td></tr>
|
||||
<tr><td><a href="https://arxiv.org/abs/2006.07733" target="_blank" rel="noopener noreferrer">BYOL</a> (<a href="https://github.com/XiaohangZhan">@XiaohangZhan</a>)</td><td>selfsup/byol/r50_bs4096_ep200.py (<a href="https://github.com/xieenze">@xieenze</a>)</td><td>default</td><td><a href="https://drive.google.com/file/d/1Whj3j5E3ShQj_VufjrJSzWiq1xcZZCXN/view?usp=sharing" target="_blank" rel="noopener noreferrer">byol_r50-e3b0c442.pth</a></td><td>4096 (256 for speed test)</td><td>200</td><td>30min57s</td></tr>
|
||||
<tr><td></td><td>selfsup/byol/r50_bs256_accumulate16_ep300.py (<a href="https://github.com/scnuhealthy">@scnuhealthy</a>)</td><td>default</td><td><a href="https://drive.google.com/file/d/12Zu9r3fE8qKF4OW6WQXa5Ec6VuA2m3j7/view?usp=sharing" target="_blank" rel="noopener noreferrer">byol_r50_bs256_accmulate16_ep300-5df46722.pth</a></td><td>256</td><td>300</td><td></td></tr>
|
||||
<tr><td></td><td>selfsup/byol/r50_bs2048_accumulate2_ep200_fp16.py (<a href="https://github.com/xieenze">@xieenze</a>)</td><td>default</td><td><a href="https://drive.google.com/file/d/1Poj-rxIebE1ykJhB8EeliSxWiCGP_rbD/view?usp=sharing" target="_blank" rel="noopener noreferrer">byol_r50_bs2048_accmulate2_ep200-e3b0c442.pth</a></td><td>256</td><td>200</td><td></td></tr>
|
||||
<table><thead><tr><th>Method (Implementator)</th><th>Config (Experimenter)</th><th>Remarks</th><th>Download link</th><th>Batch size</th><th>Epochs</th></tr></thead><tbody>
|
||||
<tr><td><a href="https://github.com/pytorch/vision/blob/master/torchvision/models/resnet.py">ImageNet</a></td><td>-</td><td>torchvision</td><td><a href="https://drive.google.com/file/d/11xA3TOcbD0qOrwpBfYonEDeseE1wMfBh/view?usp=sharing">imagenet_r50-21352794.pth</a></td><td>-</td><td>-</td></tr>
|
||||
<tr><td>Random</td><td>-</td><td>kaiming</td><td><a href="https://drive.google.com/file/d/1UaFTjd6sbKkZEE-f58Zv30bnx7C1qJBb/view?usp=sharing">random_r50-5d0fa71b.pth</a></td><td>-</td><td>-</td></tr>
|
||||
<tr><td><a href="https://www.cv-foundation.org/openaccess/content_iccv_2015/papers/Doersch_Unsupervised_Visual_Representation_ICCV_2015_paper.pdf">Relative-Loc</a> (<a href="https://github.com/Jiahao000">@Jiahao000</a>)</td><td>selfsup/relative_loc/r50.py</td><td>default</td><td><a href="https://drive.google.com/file/d/1ibk1BI3PFQxZqcxuDfHs3n7JnWKCgl8x/view?usp=sharing">relative_loc_r50-342c9097.pth</a></td><td>512</td><td>70</td></tr>
|
||||
<tr><td><a href="https://arxiv.org/abs/1803.07728">Rotation-Pred</a> (<a href="https://github.com/XiaohangZhan">@XiaohangZhan</a>)</td><td>selfsup/rotation_pred/r50.py</td><td>default</td><td><a href="https://drive.google.com/file/d/1t3oClmIvQ0p8RZ0V5yvQFltzjqBO823Y/view?usp=sharing">rotation_r50-cfab8ebb.pth</a></td><td>128</td><td>70</td></tr>
|
||||
<tr><td><a href="https://arxiv.org/abs/1807.05520">DeepCluster</a> (<a href="https://github.com/XiaohangZhan">@XiaohangZhan</a>)</td><td>selfsup/deepcluster/r50.py</td><td>default</td><td><a href="https://drive.google.com/file/d/1GxgP7pI18JtFxDIC0hnHOanvUYajoLlg/view?usp=sharing">deepcluster_r50-bb8681e2.pth</a></td><td>512</td><td>200</td></tr>
|
||||
<tr><td><a href="https://arxiv.org/abs/1805.01978">NPID</a> (<a href="https://github.com/XiaohangZhan">@XiaohangZhan</a>)</td><td>selfsup/npid/r50.py</td><td>default</td><td><a href="https://drive.google.com/file/d/1sm6I3Y5XnCWdbmeLSF4YupUtPe5nRQMI/view?usp=sharing">npid_r50-dec3df0c.pth</a></td><td>256</td><td>200</td></tr>
|
||||
<tr><td></td><td>selfsup/npid/r50_ensure_neg.py</td><td>ensure_neg=True</td><td><a href="https://drive.google.com/file/d/1FldDrb6kzF3CZ7737mwCXVI6HE2aCSaF/view?usp=sharing">npid_r50_ensure_neg-ce09b7ae.pth</a></td><td></td><td></td></tr>
|
||||
<tr><td><a href="http://openaccess.thecvf.com/content_CVPR_2020/papers/Zhan_Online_Deep_Clustering_for_Unsupervised_Representation_Learning_CVPR_2020_paper.pdf">ODC</a> (<a href="https://github.com/XiaohangZhan">@XiaohangZhan</a>)</td><td>selfsup/odc/r50_v1.py (<a href="https://github.com/Jiahao000">@Jiahao000</a>)</td><td>default</td><td><a href="https://drive.google.com/file/d/1EdhJeZAyMsD_pEW7uMhLzos5xZLdariN/view?usp=sharing">odc_r50_v1-5af5dd0c.pth</a></td><td>512</td><td>440</td></tr>
|
||||
<tr><td><a href="https://arxiv.org/abs/1911.05722">MoCo</a> (<a href="https://github.com/XiaohangZhan">@XiaohangZhan</a>)</td><td>selfsup/moco/r50_v1.py</td><td>default</td><td><a href="https://drive.google.com/file/d/1ANXfnoT8yBQQBBqR_kQLQorK20l65KMy/view?usp=sharing">moco_r50_v1-4ad89b5c.pth</a></td><td>256</td><td>200</td></tr>
|
||||
<tr><td><a href="https://arxiv.org/abs/2003.04297">MoCo v2</a> (<a href="https://github.com/XiaohangZhan">@XiaohangZhan</a>)</td><td>selfsup/moco/r50_v2.py</td><td>default</td><td><a href="https://drive.google.com/file/d/1ImO8A3uWbrTx21D1IqBDMUQvpN6wmv0d/view?usp=sharing">moco_r50_v2-e3b0c442.pth</a></td><td>256</td><td>200</td></tr>
|
||||
<tr><td><a href="https://arxiv.org/abs/2002.05709">SimCLR</a> (<a href="https://github.com/XiaohangZhan">@XiaohangZhan</a>)</td><td>selfsup/simclr/r50_bs256_ep200.py</td><td>default</td><td><a href="https://drive.google.com/file/d/1aZ43nSdivdNxHbM9DKVoZYVhZ8TNnmPp/view?usp=sharing">simclr_r50_bs256_ep200-4577e9a6.pth</a></td><td>256</td><td>200</td></tr>
|
||||
<tr><td></td><td>selfsup/simclr/r50_bs256_ep200_mocov2_neck.py</td><td>-> MoCo v2 neck</td><td><a href="https://drive.google.com/file/d/1AXpSKqgWfnj6jCgN65BXSTCKFfuIVELa/view?usp=sharing">simclr_r50_bs256_ep200_mocov2_neck-0d6e5ff2.pth</a></td><td></td><td></td></tr>
|
||||
<tr><td><a href="https://arxiv.org/abs/2006.07733">BYOL</a> (<a href="https://github.com/XiaohangZhan">@XiaohangZhan</a>)</td><td>selfsup/byol/r50_bs4096_ep200.py (<a href="https://github.com/xieenze">@xieenze</a>)</td><td>default</td><td><a href="https://drive.google.com/file/d/1Whj3j5E3ShQj_VufjrJSzWiq1xcZZCXN/view?usp=sharing">byol_r50-e3b0c442.pth</a></td><td>4096 (256 for speed test)</td><td>200</td></tr>
|
||||
<tr><td></td><td>selfsup/byol/r50_bs256_accumulate16_ep300.py (<a href="https://github.com/scnuhealthy">@scnuhealthy</a>)</td><td>default</td><td><a href="https://drive.google.com/file/d/12Zu9r3fE8qKF4OW6WQXa5Ec6VuA2m3j7/view?usp=sharing">byol_r50_bs256_accmulate16_ep300-5df46722.pth</a></td><td>256</td><td>300</td></tr>
|
||||
<tr><td></td><td>selfsup/byol/r50_bs2048_accumulate2_ep200_fp16.py (<a href="https://github.com/xieenze">@xieenze</a>)</td><td>default</td><td><a href="https://drive.google.com/file/d/1Poj-rxIebE1ykJhB8EeliSxWiCGP_rbD/view?usp=sharing">byol_r50_bs2048_accmulate2_ep200-e3b0c442.pth</a></td><td>256</td><td>200</td></tr>
|
||||
</tbody></table>
|
||||
|
||||
|
||||
## Benchmarks
|
||||
|
||||
### VOC07 SVM & SVM Low-shot
|
||||
|
|
|
|||
Loading…
Reference in New Issue