Browse › Computer Vision › Action Recognition › Something-Something V1

Action Recognition archive 2025-07-28

Something-Something V1 Benchmark (Action Recognition)

74 rows 60 with code listed 4 metrics Dataset page

Action Recognition is a computer vision task that involves recognizing human actions in videos or images. The goal is to classify and categorize the actions being performed in the video or image into a predefined set of action classes.

In the video domain, it is an open question whether training an action classification network on a sufficiently large dataset, will give a similar boost in performance when applied to a different temporal task or dataset. The challenges of building video datasets has meant that most popular benchmarks for action recognition are small, having on the order of 10k videos.

Please note some benchmarks may be located in the Action Classification or Video Classification tasks, e.g. Kinetics-400.

The archive carries no text for this table; the description above is the archive's text for the task Action Recognition. 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). Not inferred (points only, no best-so-far line): Param., GFLOPs. Points are placed at the row's paper date; 74 of 74 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 InternVideo 70.0 ✓ Paper Code 2022 3 of 3 ran · 0 unverified report
2 VideoMAE V2-g 68.791.9 ✓ Paper Code 2023 2 of 6 ran · 4 unverified report
3 Side4Video (EVA ViT-E/14 67.388.8 – Paper Code 2023 linked, not harvested report
4 ATM 65.688.6 – Paper Code 2023 3 of 6 ran · 3 unverified report
5 TAdaFormer-L/14 63.7 ✓ Paper Code 2023 2 of 2 ran · 0 unverified report
6 TDS-CLIP-ViT-L/14(8frames) 63.087.8 – Paper Code 2024 linked, not harvested report
7 UniFormerV2-L 62.788.0 ✓ Paper Code 2022 linked, not harvested report
8 StructVit-B-4-1 61.3 – Paper – 2024 no code linked report
9 UniFormer-B (IN-1K + Kinetics400) 60.987.350.1259x3 – Paper Code 2021 linked, not harvested report
10 TAdaConvNeXtV2-B 60.7 ✓ Paper Code 2023 2 of 2 ran · 0 unverified report
11 TPS 58.3 – Paper Code 2022 1 of 1 ran · 0 unverified report
12 MSMA (8+16frames) 57.9 – Paper – 2023 no code linked report
13 UniFormer-B (IN-1K + Kinetics600) 57.684.921.441.8x3 – Paper Code 2021 linked, not harvested report
14 SIFA 57.3 – Paper Code 2022 linked, not harvested report
15 EAN ResNet50 (single clip, center crop,8+16 ensemble, with sparse Transformer) 57.283.9 – Paper Code 2021 linked, not harvested report
16 TCM (Ensemble) 57.2 – Paper Code 2022 linked, not harvested report
17 BQNEn (ImageNet + K400 pretrained) 57.184.2 – Paper Code 2021 linked, not harvested report
18 TDN ResNet101 (one clip, center crop, 8+16 ensemble, ImageNet pretrained, RGB only) 56.884.1 – Paper Code 2020 2 of 4 ran · 2 unverified report
19 SELFYNet-TSM-R50En (8+16 frames, ImageNet pretrained, 2 clips) 56.684.4 ✓ Paper Code 2021 3 of 4 ran · 1 unverified report
20 CT-Net Ensemble (R50, 8+12+16+24) 56.6 – Paper Code 2021 10 of 13 ran · 3 unverified report
21 MoDS (8+16frames) 56.6 – Paper – 2022 no code linked report
22 MLP-3D 56.5 – Paper – 2022 no code linked report
23 RSANet-R50 (8+16 frames, ImageNet pretrained, 2 clips) 56.182.8 – Paper Code 2021 1 of 1 ran · 0 unverified report
24 SELFYNet-TSM-R50En (8+16 frames, ImageNet pretrained, a single clip) 55.883.9 ✓ Paper Code 2021 3 of 4 ran · 1 unverified report
25 RSANet-R50 (8+16 frames, ImageNet pretrained, a single clip) 55.582.6 – Paper Code 2021 1 of 1 ran · 0 unverified report
26 PAN ResNet101 (RGB only, no Flow) 55.382.8 – Paper Code 2020 linked, not harvested report
27 GSM Ensemble InceptionV3 (ImageNet pretrained) 55.16 ✓ Paper Code 2019 2 of 3 ran · 1 unverified report
28 MSNet-R50En (ensemble) 55.1 ✓ Paper Code 2020 3 of 8 ran · 5 unverified report
29 AE-Net (8+16frames) 55.0 – Paper – 2022 no code linked report
30 VoV3D-L (32frames, Kinetics pretrained, single) 54.5982.305.8M20.9x6 ✓ Paper Code 2020 linked, not harvested report
31 MSNet-R50En (8+16 ensemble, ImageNet pretrained) 54.483.8 ✓ Paper Code 2020 3 of 8 ran · 5 unverified report
32 SELFYNet-TSM-R50 (16 frames, ImageNet pretrained) 54.382.9 ✓ Paper Code 2021 3 of 4 ran · 1 unverified report
33 RNL+TSM Ensemble(R50+R101, ImageNet pretrained) 54.182.2 – Paper Code 2020 linked, not harvested report
34 RSANet-R50 (16 frames, ImageNet pretrained, a single clip) 54.081.1 – Paper Code 2021 1 of 1 ran · 0 unverified report
35 MVFNet-R50EN 54.0 – Paper Code 2020 linked, not harvested report
36 STPG (8+16frames) 53.5 – Paper – 2022 no code linked report
37 GB + DF + LB (ResNet152, ImageNet pretrained) 53.4 ✓ Paper – 2019 no code linked report
38 ip-CSN-152 (IG-65M pretraining) 53.3 – Paper Code 2019 1 of 4 ran · 3 unverified report
39 MARS+RGB+Flow (64 frames, Kinetics pretrained) 53.0 ✓ Paper Code 2019 linked, not harvested report
40 RNL+TSM Ensemble(ResNet50, ImageNet pretrained) 52.781.5 – Paper Code 2020 linked, not harvested report
41 VoV3D-M (32frames, Kinetics pretrained, single) 52.6880.433.3M11.5x6 ✓ Paper Code 2020 linked, not harvested report
42 TSM+W3 (16 frames, ResNet50) 52.681.3 – Paper – 2020 no code linked report
43 AK-Net 52.5 – Paper – 2022 no code linked report
44 MSNet-R50 (16 frames, ImageNet pretrained) 52.182.3 ✓ Paper Code 2020 3 of 8 ran · 5 unverified report
45 ir-CSN-152 (IG-65M pretraining) 52.1 – Paper Code 2019 1 of 4 ran · 3 unverified report
46 RSANet-R50 (8 frames, ImageNet pretrained, a single clip) 51.979.6 – Paper Code 2021 1 of 1 ran · 0 unverified report
47 GSM InceptionV3 (16 frames, ImageNet pretrained) 51.68 ✓ Paper Code 2019 2 of 3 ran · 1 unverified report
48 R(2+1)D-152 (IG-65M pretraining) 51.6 – Paper Code 2019 1 of 4 ran · 3 unverified report
49 MSNet-R50 (8 frames, ImageNet pretrained) 50.980.3 – Paper Code 2020 3 of 8 ran · 5 unverified report
50 TSM (RGB + Flow) 50.7 – Paper Code 2018 6 of 16 ran · 10 unverified report
51 VoV3D-L (32frames, from scratch, single) 50.678.75.8M20.9x6 – Paper Code 2020 linked, not harvested report
52 ResNet50 I3D (Moments pretrained) 50 ✓ Paper Code 2018 linked, not harvested report
53 VoV3D-M (32frames, from scratch, single) 49.878.03.3M11.5x6 – Paper Code 2020 linked, not harvested report
54 TSMEn 49.778.5 – Paper Code 2018 6 of 16 ran · 10 unverified report
55 TRG (Inception-V3) 49.7 – Paper – 2019 no code linked report
56 TRG (ResNet-50) 49.586.1 – Paper – 2019 no code linked report
57 VoV3D-L (16frames, from scratch, single) 49.578.05.8M9.3x6 – Paper Code 2020 linked, not harvested report
58 ir-CSN-152 49.3 – Paper Code 2019 1 of 4 ran · 3 unverified report
59 RSTG (Kinetics pretrained) 49.2 ✓ Paper Code 2019 1 of 1 ran · 0 unverified report
60 ResNet50 I3D (Kinetics pretrained) 48.6 ✓ Paper Code 2018 linked, not harvested report
61 ir-CSN-101 48.4 – Paper Code 2019 1 of 4 ran · 3 unverified report
62 S3D-G (ImageNet pretrained) 48.278.7 ✓ Paper Code 2017 linked, not harvested report
63 VoV3D-M (16frames, from scratch, single) 48.176.93.3M5.7x6 – Paper Code 2020 linked, not harvested report
64 S3D 47.378.1 – Paper Code 2017 linked, not harvested report
65 TSM 47.277.1 – Paper Code 2018 6 of 16 ran · 10 unverified report
66 ECO-Net (ImageNet pretrained) 46.4 ✓ Paper Code 2018 0 of 2 ran · 2 unverified report
67 ECO-Net 46.4 – Paper Code 2018 0 of 2 ran · 2 unverified report
68 NL I3D + GCN 46.1 – Paper – 2018 no code linked report
69 NL I3D 44.4 – Paper Code 2017 3 of 4 ran · 1 unverified report
70 Motion Feature Net 43.9 – Paper – 2018 no code linked report
71 2-Stream TRN 42.01 – Paper Code 2017 2 of 3 ran · 1 unverified report
72 HF-TSN (ImageNet pretraining) 41.97 ✓ Paper – 2019 no code linked report
73 MARS+RGB+Flow (16 frames, Kinetics pretrained) 40.4 – Paper Code 2019 linked, not harvested report
74 M-TRN 34.4 – Paper Code 2017 2 of 3 ran · 1 unverified report

All 74 rows shown. 74 link to a paper page on this site; 23 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). 35 rows have a graph line, from 17 distinct papers; 33 rows (16 papers) have at least one sample that ran. Counting each paper once: Syntology ran 45 of 81 samples; 36 unverified. Separately, 19 of those 81 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