Hugging Face's logo

lightly-ai
/
in1k-benchmark-logs

TensorBoard
Model card Files Metrics Community
SauravMaheshkar commited on
Commit
3955a4b
•
1 Parent(s): 7408887

docs: update README

Browse files
Files changed (1) hide show
  1. README.md +32 -0
README.md CHANGED
@@ -1,3 +1,35 @@
1
  ---
2
  license: mit
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: mit
3
  ---
4
+
5
+ # Lightly ImageNet1k Benchmarks
6
+
7
+ ![LightlySSL self-supervised learning Logo](https://github.com/lightly-ai/lightly/blob/master/docs/logos/lightly_SSL_logo_crop.png?raw=true)
8
+
9
+ ![GitHub](https://img.shields.io/github/license/lightly-ai/lightly)
10
+ [![Discord](https://img.shields.io/discord/752876370337726585?logo=discord&logoColor=white&label=discord&color=7289da)](https://discord.gg/xvNJW94)
11
+
12
+ **Note**: Evaluation settings are based on these papers:
13
+
14
+ - Linear: [SimCLR](https://arxiv.org/abs/2002.05709)
15
+ - Finetune: [SimCLR](https://arxiv.org/abs/2002.05709)
16
+ - KNN: [InstDisc](https://arxiv.org/abs/1805.01978)
17
+
18
+ See the [benchmarking scripts](./benchmarks/imagenet/resnet50/) for details.
19
+
20
+ | Model | Backbone | Batch Size | Epochs | Linear Top1 | Finetune Top1 | kNN Top1 | Tensorboard | Checkpoint |
21
+ | --------------- | -------- | ---------- | ------ | ----------- | ------------- | -------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
22
+ | BarlowTwins | Res50 | 256 | 100 | 62.9 | 72.6 | 45.6 | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_barlowtwins_2023-08-18_00-11-03/pretrain/version_0/events.out.tfevents.1692310273.Machine2.569794.0) | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_barlowtwins_2023-08-18_00-11-03/pretrain/version_0/checkpoints/epoch%3D99-step%3D500400.ckpt) |
23
+ | BYOL | Res50 | 256 | 100 | 62.5 | 74.5 | 46.0 | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_byol_2024-02-14_16-10-09/pretrain/version_0/events.out.tfevents.1707923418.Machine2.3205.0) | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_byol_2024-02-14_16-10-09/pretrain/version_0/checkpoints/epoch%3D99-step%3D500400.ckpt) |
24
+ | DINO | Res50 | 128 | 100 | 68.2 | 72.5 | 49.9 | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_dino_2023-06-06_13-59-48/pretrain/version_0/events.out.tfevents.1686052799.Machine2.482599.0) | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_dino_2023-06-06_13-59-48/pretrain/version_0/checkpoints/epoch%3D99-step%3D1000900.ckpt) |
25
+ | MAE | ViT-B/16 | 256 | 100 | 46.0 | 81.3 | 11.2 | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_vitb16_mae_2024-02-25_19-57-30/pretrain/version_0/events.out.tfevents.1708887459.Machine2.1092409.0) | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_vitb16_mae_2024-02-25_19-57-30/pretrain/version_0/checkpoints/epoch%3D99-step%3D500400.ckpt) |
26
+ | MoCoV2 | Res50 | 256 | 100 | 61.5 | 74.3 | 41.8 | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_mocov2_2024-02-18_10-29-14/pretrain/version_0/events.out.tfevents.1708248562.Machine2.439033.0) | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_mocov2_2024-02-18_10-29-14/pretrain/version_0/checkpoints/epoch%3D99-step%3D500400.ckpt) |
27
+ | SimCLR\* | Res50 | 256 | 100 | 63.2 | 73.9 | 44.8 | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_simclr_2023-06-22_09-11-13/pretrain/version_0/events.out.tfevents.1687417883.Machine2.33270.0) | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_simclr_2023-06-22_09-11-13/pretrain/version_0/checkpoints/epoch%3D99-step%3D500400.ckpt) |
28
+ | SimCLR\* + DCL | Res50 | 256 | 100 | 65.1 | 73.5 | 49.6 | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_dcl_2023-07-04_16-51-40/pretrain/version_0/events.out.tfevents.1688482310.Machine2.247807.0) | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_dcl_2023-07-04_16-51-40/pretrain/version_0/checkpoints/epoch%3D99-step%3D500400.ckpt) |
29
+ | SimCLR\* + DCLW | Res50 | 256 | 100 | 64.5 | 73.2 | 48.5 | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_dclw_2023-07-07_14-57-13/pretrain/version_0/events.out.tfevents.1688734645.Machine2.3176.0) | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_dclw_2023-07-07_14-57-13/pretrain/version_0/checkpoints/epoch%3D99-step%3D500400.ckpt) |
30
+ | SwAV | Res50 | 256 | 100 | 67.2 | 75.4 | 49.5 | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_swav_2023-05-25_08-29-14/pretrain/version_0/events.out.tfevents.1684996168.Machine2.1445108.0) | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_swav_2023-05-25_08-29-14/pretrain/version_0/checkpoints/epoch%3D99-step%3D500400.ckpt) |
31
+ | TiCo | Res50 | 256 | 100 | 49.7 | 72.7 | 26.6 | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_tico_2024-01-07_18-40-57/pretrain/version_0/events.out.tfevents.1704649265.Machine2.1604956.0) | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_tico_2024-01-07_18-40-57/pretrain/version_0/checkpoints/epoch%3D99-step%3D250200.ckpt) |
32
+ | VICReg | Res50 | 256 | 100 | 63.0 | 73.7 | 46.3 | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_vicreg_2023-09-11_10-53-08/pretrain/version_0/events.out.tfevents.1694422401.Machine2.556563.0) | [link](https://lightly-ssl-checkpoints.s3.amazonaws.com/imagenet_resnet50_vicreg_2023-09-11_10-53-08/pretrain/version_0/checkpoints/epoch%3D99-step%3D500400.ckpt) |
33
+
34
+ _\*We use square root learning rate scaling instead of linear scaling as it yields
35
+ better results for smaller batch sizes. See Appendix B.1 in the [SimCLR paper](https://arxiv.org/abs/2002.05709)._