</> Code with Papers by Syntology

Browse › Self-Supervised Image Classification › ImageNet

Self-Supervised Image Classification archive 2025-07-28

ImageNet Benchmark (Self-Supervised Image Classification)

144 rows 135 with code listed 3 metrics Dataset page

The archive's category list for this table covers 11 areas (Adversarial, Audio, Computer Code, Computer Vision, Medical, Methodology, Miscellaneous, Music, Natural Language Processing, Reasoning, Speech), which is its placeholder rather than an assignment; no area is shown in the breadcrumb. archive 2025-07-28

This is the task of image classification using representations learnt with self-supervised learning. Self-supervised methods generally involve a pretext task that is solved to learn a good representation and a loss function to learn with. One example of a loss function is an autoencoder based loss where the goal is reconstruction of an image pixel-by-pixel. A more popular recent example is a contrastive loss, which measure the similarity of sample pairs in a representation space, and where there can be a varying target instead of a fixed target to reconstruct (as in the case of autoencoders).

A common evaluation protocol is to train a linear classifier on top of (frozen) representations learnt by self-supervised methods. The leaderboards for the linear evaluation protocol can be found below. In practice, it is more common to fine-tune features on a downstream task. An alternative evaluation protocol therefore uses semi-supervised learning and finetunes on a % of the labels. The leaderboards for the finetuning protocol can be accessed here.

You may want to read some blog posts before reading the papers and checking the leaderboards:

There is also Yann LeCun's talk at AAAI-20 which you can watch here (35:00+).

The archive carries no text for this table; the description above is the archive's text for the task Self-Supervised Image Classification. archive 2025-07-28

Over time archive 2025-07-28

The chart needs JavaScript; the table below carries every value.

Direction inferred from the metric name, not from the archive: Top 1 Accuracy (higher is better), Top 5 Accuracy (higher is better), Number of Params (lower is better). Points are placed at the row's paper date; 144 of 144 rows carry one.

Results archive 2025-07-28

Archive rows end at the archive snapshot, 2025-07-28: no result published after that date is in this table. Rank is the archive's row order at that snapshot; not re-ranked here. Metric values are the archive's strings. Column headers sort the table in your browser; each row keeps its archive rank.

Paper Code Ran Syntology Report
1 DINOv2+reg (ViT-g/14) 87.11100M ✓ Paper Code 2023 4 of 20 ran · 16 unverified report
2 DINOv2 (ViT-g/14 @448) 86.7%1100M ✓ Paper Code 2023 21 of 46 ran · 25 unverified report
3 DINOv2 (ViT-g/14) 86.5%1100M ✓ Paper Code 2023 21 of 46 ran · 25 unverified report
4 DINOv2 distilled (ViT-L/14) 86.3%307M ✓ Paper Code 2023 21 of 46 ran · 25 unverified report
5 MIM-Refiner (D2V2-ViT-H/14) 84.7%632M – Paper Code 2024 3 of 3 ran · 0 unverified report
6 MIM-Refiner (MAE-ViT-2B/14) 84.5%1890M – Paper Code 2024 3 of 3 ran · 0 unverified report
7 DINOv2 distilled (ViT-B/14) 84.5%85M ✓ Paper Code 2023 21 of 46 ran · 25 unverified report
8 MIM-Refiner (MAE-ViT-H/14 83.7%632M – Paper Code 2024 3 of 3 ran · 0 unverified report
9 MIM-Refiner (D2V2-ViT-L/16) 83.5%307M – Paper Code 2024 3 of 3 ran · 0 unverified report
10 MIM-Refiner (MAE-ViT-L/16) 82.8%307M – Paper Code 2024 3 of 3 ran · 0 unverified report
11 iBOT (ViT-L/16) (IN22k) 82.3%307M ✓ Paper Code 2021 0 of 1 ran · 1 unverified report
12 MAE-CT (ViT-H/16) 82.2%632M – Paper Code 2023 linked, not harvested report
13 Mugs (VIT-L/16) 82.1%307M – Paper Code 2022 4 of 21 ran · 17 unverified report
14 MAE-CT (ViT-L/16 81.5%307M – Paper Code 2023 linked, not harvested report
15 EsViT (Swin-B) 81.395.587M – Paper Code 2021 2 of 6 ran · 4 unverified report
16 iBOT (ViT-L/16) 81.3%307M – Paper Code 2021 0 of 1 ran · 1 unverified report
17 DINOv2 distilled (ViT-S/14) 81.1%21M ✓ Paper Code 2023 21 of 46 ran · 25 unverified report
18 MoCo v3 (ViT-BN-L/7) 81.0%304M – Paper Code 2021 2 of 2 ran · 0 unverified report
19 EsViT(Swin-S) 80.849M – Paper Code 2021 2 of 6 ran · 4 unverified report
20 MSN (ViT-L/7) 80.7%306M – Paper Code 2022 4 of 4 ran · 0 unverified report
21 ReLICv2 (ResNet-200 x2) 80.6%250M – Paper Code 2022 0 of 14 ran · 14 unverified report
22 MR BarTwins (MR BarTwins) 80.4% – Paper – 2022 no code linked report
23 DiGIT 80.3%732M – Paper Code 2024 3 of 5 ran · 2 unverified report
24 iBOT-vMF (ViT-B/16) 80.3%85M – Paper – 2024 no code linked report
25 DINO (xcit_medium_24_p8) 80.3%84M – Paper Code 2021 5 of 20 ran · 15 unverified report
26 PGT (PGT-B w/ Flow) 80.3%70M – Paper – 2023 no code linked report
27 DINO (ViT-B/8) 80.1%80M – Paper Code 2021 5 of 20 ran · 15 unverified report
28 SimCLRv2 (ResNet-152 x3, SK) 79.8%94.9%795M – Paper Code 2020 0 of 6 ran · 6 unverified report
29 SEERv2 79.8%10000M ✓ Paper Code 2022 linked, not harvested report
30 PercMAE (ViT-B, dVAE) 79.8%80M ✓ Paper Code 2022 linked, not harvested report
31 ReLICv2 (ResNet200) 79.8%63M – Paper Code 2022 0 of 14 ran · 14 unverified report
32 DINO (ViT-S/8) 79.7%21M – Paper Code 2021 5 of 20 ran · 15 unverified report
33 BYOL (ResNet-200 x2) 79.6%94.8%250M – Paper Code 2020 62 of 79 ran · 17 unverified report
34 ReLICv2 (ResNet-50 4x) 79.4%375M – Paper Code 2022 0 of 14 ran · 14 unverified report
35 ReLICv2 (ResNet152) 79.3%58M – Paper Code 2022 0 of 14 ran · 14 unverified report
36 BAM (CAFormer-M36) 79.3% – Paper Code 2024 linked, not harvested report
37 MoCo v3 (ViT-BN-H) 79.1%700M – Paper Code 2021 2 of 2 ran · 0 unverified report
38 Unicom (ViT-B/16) 79.1%80M ✓ Paper Code 2023 3 of 6 ran · 3 unverified report
39 SMoG (ResNet-50 x4) 79.0%94.4375M – Paper Code 2022 1 of 1 ran · 0 unverified report
40 ReLICv2 (ResNet-50 x2) 79%94M – Paper Code 2022 0 of 14 ran · 14 unverified report
41 C-BYOL (ResNet-50 2x, 1000 epochs) 78.8%94.5%94M – Paper Code 2021 0 of 12 ran · 12 unverified report
42 DINO-vMF (ViT-B/16) 78.8%85M – Paper – 2024 no code linked report
43 ReLICv2 (ResNet101) 78.7%44M – Paper Code 2022 0 of 14 ran · 14 unverified report
44 BYOL (ResNet-50 x4) 78.6%94.2%375M – Paper Code 2020 62 of 79 ran · 17 unverified report
45 SwAV (ResNet-50 x5) 78.5%586M – Paper Code 2020 13 of 17 ran · 4 unverified report
46 DINO (ViT-B/16) 78.2%85M – Paper Code 2021 5 of 20 ran · 15 unverified report
47 MoCo v3 (ViT-H) 78.1%632M – Paper Code 2021 2 of 2 ran · 0 unverified report
48 PercMAE (ViT-B) 78.1%80M – Paper Code 2022 linked, not harvested report
49 BAM (ViT-B/16) 78.1%80M – Paper Code 2024 linked, not harvested report
50 SMoG (ResNet-50 x2) 78.0%93.994M – Paper Code 2022 1 of 1 ran · 0 unverified report
51 MoCo v3 (ViT-L) 77.6%307M – Paper Code 2021 2 of 2 ran · 0 unverified report
52 SEER 77.5%1300M ✓ Paper Code 2021 linked, not harvested report
53 BYOL (ResNet-50 x2) 77.4%93.6%94M – Paper Code 2020 62 of 79 ran · 17 unverified report
54 SwAV (ResNet-50 x2) 77.3%94M – Paper Code 2020 13 of 17 ran · 4 unverified report
55 ReLICv2 (ResNet-50) 77.1%25M – Paper Code 2022 0 of 14 ran · 14 unverified report
56 DINO (ViT-S/16) 77.0%21M – Paper Code 2021 5 of 20 ran · 15 unverified report
57 DINO-vMF (ViT-S/16) 77.0%21M – Paper – 2024 no code linked report
58 MoCo v3 (ViT-B/16) 76.7%86M – Paper Code 2021 2 of 2 ran · 0 unverified report
59 MAE (ViT-H) 76.6%700M – Paper Code 2021 71 of 137 ran · 66 unverified report
60 SimCLR (ResNet-50 4x) 76.5%93.2%375M – Paper Code 2020 79 of 137 ran · 58 unverified report
61 CoKe (ResNet-50) 76.4%25M – Paper Code 2021 3 of 6 ran · 3 unverified report
62 SMoG (ResNet-50) 76.4%25M – Paper Code 2022 1 of 1 ran · 0 unverified report
63 ReSSL (ResNet-50 w/ Predictor and Stronger Aug) 76.3%24M – Paper Code 2022 0 of 2 ran · 2 unverified report
64 ReSSL (ResNet-50 w/ Predictor) 76.0%24M – Paper Code 2022 0 of 2 ran · 2 unverified report
65 Triplet (ResNet-50) 75.9%23.56M – Paper Code 2021 0 of 2 ran · 2 unverified report
66 MAE (ViT-L) 75.8%306M – Paper Code 2021 71 of 137 ran · 66 unverified report
67 DnC (ResNet-50) 75.8%24M – Paper – 2021 no code linked report
68 CaCo (ResNet-50) 75.7%24M – Paper Code 2022 2 of 4 ran · 2 unverified report
69 SimCLRv2 (ResNet-50 x2) 75.6%92.7%94M – Paper Code 2020 0 of 6 ran · 6 unverified report
70 C-BYOL (ResNet-50, 1000 epochs) 75.6%92.7%25M – Paper Code 2021 0 of 12 ran · 12 unverified report
71 NNCLR (ResNet-50, multi-crop) 75.6%92.425M – Paper Code 2021 4 of 5 ran · 1 unverified report
72 HEXA 75.5%24M – Paper – 2020 no code linked report
73 SCE (ResNet-50, multi-crop) 75.4%24M – Paper Code 2021 linked, not harvested report
74 SwAV (ResNet-50) 75.3%24M – Paper Code 2020 13 of 17 ran · 4 unverified report
75 DINO (ResNet-50) 75.3%24M – Paper Code 2021 5 of 20 ran · 15 unverified report
76 InfoMin (ResNeXt-152) 75.2%120M – Paper Code 2020 linked, not harvested report
77 DeepCluster-v2 (ResNet-50) 75.2%24M – Paper Code 2020 13 of 17 ran · 4 unverified report
78 Unicom (ViT-B/32) 75.0%80M ✓ Paper Code 2023 3 of 6 ran · 3 unverified report
79 MoBY (Swin-T) 75%29M – Paper Code 2021 linked, not harvested report
80 ReLIC (ResNet-50) 74.8%24M – Paper Code 2020 2 of 2 ran · 0 unverified report
81 ReSSL(ResNet-50) 200ep 74.7%92.3%24M – Paper Code 2021 1 of 1 ran · 0 unverified report
82 WCL (ResNet-50) 74.7%24M – Paper Code 2021 0 of 3 ran · 3 unverified report
83 MV-MR 74.5%92.1 – Paper Code 2023 linked, not harvested report
84 FNC (ResNet-50) 74.4%91.8%24M – Paper Code 2020 linked, not harvested report
85 BYOL (ResNet-50) 74.3%91.6%24M – Paper Code 2020 62 of 79 ran · 17 unverified report
86 SimCLR (ResNet-50 2x) 74.2%92.0%94M – Paper Code 2020 79 of 137 ran · 58 unverified report
87 Self-Classifier (ResNet-50) 74.2%24M – Paper Code 2021 linked, not harvested report
88 GroCo (ResNet-50) 73.9%91.625M – Paper Code 2023 linked, not harvested report
89 OBoW (ResNet-50) 73.8%92.2%24M – Paper Code 2020 3 of 3 ran · 0 unverified report
90 VICReg (ResNet50) 73.291.124M – Paper Code 2021 11 of 17 ran · 6 unverified report
91 Barlow Twins (ResNet-50) 73.2%9124M – Paper Code 2021 21 of 26 ran · 5 unverified report
92 InfoMin (ResNet-50) 73.0%91.1%24M – Paper Code 2020 linked, not harvested report
93 DINO (ResMLP-24) 72.8%30M – Paper Code 2021 2 of 7 ran · 5 unverified report
94 MoBY (DeiT-S) 72.8%22M – Paper Code 2021 linked, not harvested report
95 I-VNE+ (ResNet-50) 72.191.025M – Paper Code 2023 linked, not harvested report
96 iGPT-XL (64x64, 15360 features) 72.0%6801M – Paper Code 2020 linked, not harvested report
97 SimCLRv2 (ResNet-50) 71.7%90.4%24M – Paper Code 2020 0 of 6 ran · 6 unverified report
98 CPC v2 (ResNet-161) (arxiv v2) 71.5%90.1%305M – Paper Code 2019 2 of 2 ran · 0 unverified report
99 SimSiam (ResNet-50) 71.3%24M – Paper Code 2020 34 of 39 ran · 5 unverified report
100 MoCo v2 (ResNet-50) 71.1%90.1%24M – Paper Code 2020 8 of 43 ran · 35 unverified report
101 SynCo (ResNet-50) 800ep 70.6%89.8%24M – Paper Code 2024 linked, not harvested report
102 CMC (ResNet-50 x2) (arxiv v5) 70.6%89.7%188M – Paper Code 2019 3 of 4 ran · 1 unverified report
103 SimCLR (ResNet-50) 69.3%89.0%24M – Paper Code 2020 79 of 137 ran · 58 unverified report
104 iGPT-XL (64x64, 3072 features) 68.7%6800M – Paper Code 2020 linked, not harvested report
105 MoCo (ResNet-50 4x) 68.6%375M – Paper Code 2019 26 of 42 ran · 16 unverified report
106 AMDIM (large) (arxiv v2) 68.1%626M – Paper Code 2019 8 of 11 ran · 3 unverified report
107 MAE (ViT-B) 68.0%80M – Paper Code 2021 71 of 137 ran · 66 unverified report
108 SynCo (ResNet-50) 200ep 67.9%8824M – Paper Code 2024 linked, not harvested report
109 DINO (ResMLP-12) 67.5%15M – Paper Code 2021 2 of 7 ran · 5 unverified report
110 CMC (ResNet-50) (arxiv v5) 66.2%87.0%47M – Paper Code 2019 3 of 4 ran · 1 unverified report
111 PCL (ResNet-50) 65.9%25M – Paper Code 2020 2 of 5 ran · 3 unverified report
112 MoCo (ResNet-50 2x) 65.4%94M – Paper Code 2019 26 of 42 ran · 16 unverified report
113 iGPT-L (48x48) 65.2%1400M – Paper Code 2020 linked, not harvested report
114 CMC (ResNet-101) (arxiv v3) 65.0%86.0% – Paper Code 2019 3 of 4 ran · 1 unverified report
115 CPC v2 (ResNet-50) (arxiv v2) 63.8%85.3%24M – Paper Code 2019 2 of 2 ran · 0 unverified report
116 MMCL (100 epoch, 256 batch size) 63.8% – Paper Code 2021 2 of 12 ran · 10 unverified report
117 PIRL 63.6%24M – Paper Code 2019 3 of 4 ran · 1 unverified report
118 AMDIM (small) (arxiv v2) 63.5%194M – Paper Code 2019 8 of 11 ran · 3 unverified report
119 SeLa (ResNet50) (arxiv 3) 61.5%84.0%24M – Paper Code 2019 7 of 18 ran · 11 unverified report
120 BigBiGAN (RevNet-50 ×4, BN+CReLU) 61.3%81.9%86M – Paper Code 2019 0 of 3 ran · 3 unverified report
121 CPC v2 (ResNet-161) (arxiv v1) 61.0%83.0%305M – Paper Code 2019 2 of 2 ran · 0 unverified report
122 BigBiGAN (RevNet-50 ×4) 60.8%81.4%86M – Paper Code 2019 0 of 3 ran · 3 unverified report
123 MoCo (ResNet-50) 60.6%24M – Paper Code 2019 26 of 42 ran · 16 unverified report
124 iGPT-L (32x32) 60.3%1400M – Paper Code 2020 linked, not harvested report
125 AMDIM (arxiv v1) 60.2%337M – Paper Code 2019 8 of 11 ran · 3 unverified report
126 LocalAgg (ResNet-50) 60.2%24M – Paper Code 2019 linked, not harvested report
127 CMC (ResNet-101) 60.1%82.8%44M – Paper Code 2019 3 of 4 ran · 1 unverified report
128 BigBiGAN (ResNet-50, BN+CReLU) 56.6%78.6%24M – Paper Code 2019 0 of 3 ran · 3 unverified report
129 SeLa (ResNet50) 55.7%79.5%24M – Paper Code 2019 7 of 18 ran · 11 unverified report
130 Revisited Rotation (RevNet-50 ×4) 55.4%77.9%86M – Paper Code 2019 0 of 15 ran · 15 unverified report
131 BigBiGAN (ResNet-50) 55.4%77.4%25M – Paper Code 2019 0 of 3 ran · 3 unverified report
132 Revisited Rel.Patch.Loc (ResNet50 ×2) 51.4%74.0%94M – Paper Code 2019 0 of 15 ran · 15 unverified report
133 SeLa (AlexNet) (arxiv v3) 50.0%61M – Paper Code 2019 7 of 18 ran · 11 unverified report
134 CPC (ResNet-101 V2) 48.7%73.6%44M – Paper Code 2018 29 of 45 ran · 16 unverified report
135 Revisited Exemplar (ResNet-50 ×3) 46.0%68.8%211M – Paper Code 2019 0 of 15 ran · 15 unverified report
136 Revisited Jigsaw (ResNet50 ×2) 44.6%68.0%94M – Paper Code 2019 0 of 15 ran · 15 unverified report
137 CMC (Alexnet/2) 42.6%30M – Paper Code 2019 3 of 4 ran · 1 unverified report
138 DeepCluster (AlexNet) 41.061M – Paper Code 2018 5 of 7 ran · 2 unverified report
139 Colorisation (improved) (ResNet-101) 39.662.544M – Paper – 2017 no code linked report
140 Rotation (AlexNet) 38.786M – Paper Code 2018 14 of 22 ran · 8 unverified report
141 Split-Brain (AlexNet) 35.4%61M – Paper Code 2016 linked, not harvested report
142 Counting (AlexNet) 34.361M – Paper Code 2017 1 of 1 ran · 0 unverified report
143 Colorization (AlexNet) 32.6%61M – Paper Code 2016 32 of 73 ran · 41 unverified report
144 Multi-task SSL (ResNet-101) 70.244M – Paper – 2017 no code linked report

All 144 rows shown. 144 link to a paper page on this site; 12 are marked as using additional training data in the archive. No GitHub stars are tracked; "Code" is the first repository the archive lists for the row. The archive carries no row tags, review links or community-submitted rows for this table; none are shown. archive 2025-07-28

Syntology Ran reads "N of M ran · U unverified": of the M code samples Syntology harvested from repositories linked to that row's paper (joined by arXiv id), N executed on a synthesized input and the other U = M−N are unverified (harvested, no recorded run). It counts code from repositories linked to that row's paper, not this result: the row's number was not reproduced and nothing here is a correctness claim. The other cell texts mean no graph line for the row: "linked, not harvested" (the archive links code, Syntology has not harvested it), "no code linked" (no code link in the archive), "not matched" (the row's paper URL matched no paper on this site). 109 rows have a graph line, from 48 distinct papers; 83 rows (39 papers) have at least one sample that ran. Counting each paper once: Syntology ran 502 of 961 samples; 459 unverified. Separately, 377 of those 961 are pointer-only (licence): the site points at that code rather than redistributing it, a licence property recorded for ran and unverified samples alike; each cell's tooltip carries the row's own pointer-only count. Read from the graph 2026-09-24. Per-sample status is on the paper page.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections