Temporal Action Segmentation on Breakfast (F1@10/25/50, Edit, Acc)
76.4AccuracyDiffAct
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DiffActParams(M)=1.2, Backbone=I3D2024.09 | 76.4 | 80.3 | 75.9 | 64.6 | 78.4 | — | |
| C2F-TCNArchitecture Type=Complex, Base Backbone Model=C2F-TCN2026.04 | 76.1 | 71.41 | 67.8 | 57.28 | 67.84 | — | |
| FACT + OursArchitecture Type=Complex, Base Backbone Model=FACT2026.04 | 76.08 | 81.36 | 76.49 | 65.98 | 79.44 | — | |
| C2F-TCN + OursArchitecture Type=Complex, Base Backbone Model=C2F-TCN2026.04 | 76.03 | 72.4 | 68.59 | 57.69 | 68.54 | — | |
| Temporal AggregatesBackbone=I3D, Evaluation protocol=GT segments (Upper bound)2020.06 | 75.9 | — | — | — | — | — | |
| FACTArchitecture Type=Complex, Base Backbone Model=FACT2026.04 | 75.89 | 80.36 | 75.79 | 65.47 | 79.23 | — | |
| DiffActInference mode=offline2024.11 | 75.1 | 80.3 | 75.9 | 75.1 | 76.4 | — | |
| DiffActParams(M)=1.2, Backbone=ViT-B2024.09 | 74.9 | 78.7 | 74.9 | 62.7 | 76.4 | — | |
| OursSupervision Setting=Fully-supervised, Backbone=ASFormer2026.05 | 74.6 | 78.8 | 73.1 | 58 | 77.9 | — | |
| KARI*Supervision Setting=Fully-supervised, Backbone=ASFormer2026.05 | 74.5 | 77.3 | 71.6 | 57 | 78.3 | — | |
| Baseline [45]reproduced=false, refinement=false2023.12 | 74.3 | 77.3 | 72 | 59.4 | 75.6 | — | |
| ASFormerre-produced=true, refinement=false, grammar induction=None2023.12 | 74.3 | 77.3 | 72 | 59.4 | 75.6 | — | |
| ASFormerre-prod.=true, refinement=false, grammar induction=none2023.12 | 74.3 | 77.3 | 72 | 59.4 | 75.6 | — | |
| Richard et al.reproduced=true, refinement=true2023.12 | 74.1 | 77.2 | 72.2 | 59.4 | 77.3 | — | |
| KARIreproduced=true, refinement=true2023.12 | 74 | 78.8 | 73.7 | 60.8 | 77.8 | — | |
| ASFormer + KARIre-produced=true, refinement=true, grammar induction=KARI2023.12 | 74 | 78.8 | 73.7 | 60.8 | 77.8 | — | |
| ASFormerre-prod.=true, refinement=true, grammar induction=KARI2023.12 | 74 | 78.8 | 73.7 | 60.8 | 77.8 | — | |
| ASFormerre-produced=false, refinement=false, grammar induction=None2023.12 | 73.5 | 76 | 70.6 | 57.4 | 75 | — | |
| ASFormerre-prod.=false, refinement=false, grammar induction=none2023.12 | 73.5 | 76 | 70.6 | 57.4 | 75 | — | |
| ASFormerInference mode=offline2024.11 | 73.5 | 76 | 70.6 | 57.4 | 75 | — | |
| OriginalBackbone=ASFormer2025.03 | 73.5 | 76 | 70.6 | 57.4 | 75 | — | |
| ASFormerParams(M)=1.1, Backbone=I3D2024.09 | 73.5 | 76 | 70.6 | 57.4 | 75 | — | |
| Kuehne et al.reproduced=false, refinement=true2023.12 | 72.9 | 74 | 69.1 | 55.5 | 72.8 | — | |
| ADIOS-ANDreproduced=false, refinement=true2023.12 | 72.4 | 69.8 | 64.9 | 52.2 | 69.2 | — | |
| ASFormer + ADIOS-ANDre-produced=true, refinement=true, grammar induction=ADIOS-AND2023.12 | 72.4 | 69.8 | 64.9 | 52.2 | 69.2 | — | |
| ASFormerre-prod.=true, refinement=true, grammar induction=ADIOS-AND2023.12 | 72.4 | 69.8 | 64.9 | 52.2 | 69.2 | — | |
| w/o constraintsSupervision Setting=Fully-supervised, Backbone=ASFormer2026.05 | 72.4 | 74.1 | 68.7 | 55.5 | 72.8 | — | |
| Temporal2Seq_batchParams(M)=6.9, Backbone=ViT-B2024.09 | 72.2 | 77.5 | 71.6 | 59.7 | 75 | — | |
| ADIOS-ORreproduced=true, refinement=true2023.12 | 71.8 | 71.8 | 66.8 | 54.2 | 70.3 | — | |
| ASFormer + ADIOS-ORre-produced=true, refinement=true, grammar induction=ADIOS-OR2023.12 | 71.8 | 71.8 | 66.8 | 54.2 | 70.3 | — | |
| ASFormerre-prod.=true, refinement=true, grammar induction=ADIOS-OR2023.12 | 71.8 | 71.8 | 66.8 | 54.2 | 70.3 | — | |
| Temporal2Seq_dataParams(M)=6.9, Backbone=ViT-B2024.09 | 71.8 | 77 | 71.5 | 60.3 | 74.7 | — | |
| Baseline_TASParams(M)=6.9, Backbone=ViT-B2024.09 | 70.7 | 75.7 | 70.8 | 57.8 | 73.4 | — | |
| BCN2020.12 | 70.4 | 68.7 | 65.5 | 55 | 66.2 | — | |
| BCNSupervision=Full2021.03 | 70.4 | 68.7 | 65.5 | 55 | 66.2 | — | |
| SSTDALabeled training data percentage=100%2020.03 | 70.2 | 75 | 69.1 | 55.2 | 73.7 | — | |
| UVASTParams(M)=1.3, Backbone=I3D2024.09 | 69.7 | 76.9 | 71.5 | 58 | 77.1 | — | |
| MS-TCNre-produced=true, refinement=false, grammar induction=None2023.12 | 69.4 | 70.7 | 65.1 | 52.6 | 69.7 | — | |
| MS-TCNre-prod.=true, refinement=false, grammar induction=none2023.12 | 69.4 | 70.7 | 65.1 | 52.6 | 69.7 | — | |
| HASRParams(M)=19.2, Backbone=I3D2024.09 | 69.4 | 74.7 | 69.5 | 57 | 71.9 | — | |
| MS-TCNInference mode=offline2024.11 | 69.3 | 64.7 | 59.6 | 47.5 | 67.3 | — | |
| MS-TCN + KARIre-produced=true, refinement=true, grammar induction=KARI2023.12 | 68.8 | 74.6 | 68.7 | 55.1 | 74.9 | — | |
| MS-TCNre-prod.=true, refinement=true, grammar induction=KARI2023.12 | 68.8 | 74.6 | 68.7 | 55.1 | 74.9 | — | |
| Timestamp Supervision approachFraction=0.12021.03 | 68.4 | — | — | — | — | — | |
| MS-TCN + ADIOS-ANDre-produced=true, refinement=true, grammar induction=ADIOS-AND2023.12 | 68.4 | 66.7 | 61 | 48 | 68 | — | |
| MS-TCNre-prod.=true, refinement=true, grammar induction=ADIOS-AND2023.12 | 68.4 | 66.7 | 61 | 48 | 68 | — | |
| DTGRM2020.12 | 68.3 | 68.7 | 61.9 | 46.6 | 68.9 | — | |
| MS-TCN + ADIOS-ORre-produced=true, refinement=true, grammar induction=ADIOS-OR2023.12 | 68.2 | 69.2 | 63.3 | 50.3 | 69.6 | — | |
| MS-TCNre-prod.=true, refinement=true, grammar induction=ADIOS-OR2023.12 | 68.2 | 69.2 | 63.3 | 50.3 | 69.6 | — | |
| UARLParams(M)=1.5, Backbone=I3D2024.09 | 67.8 | 65.2 | 59.4 | 47.4 | 66.2 | — | |
| MSTCN++2020.12 | 67.6 | 64.1 | 58.6 | 45.9 | 65.6 | — | |
| MS-TCN++Supervision=Full2021.03 | 67.6 | 64.1 | 58.6 | 45.9 | 65.6 | — | |
| ASRFSupervision=Full2021.03 | 67.6 | 74.3 | 68.9 | 56.1 | 72.4 | — | |
| MS-TCN++2026.02 | 67.6 | 64.1 | 58.6 | 45.9 | 65.6 | — | |
| MS-TCN++Params(M)=0.8, Backbone=I3D2024.09 | 67.6 | 64.1 | 58.6 | 45.9 | 65.6 | — | |
| ASRFParams(M)=1.2, Backbone=I3D2024.09 | 67.6 | 74.3 | 68.9 | 56.1 | 72.4 | — | |
| MS-TCN + OursArchitecture Type=Standard, Base Backbone Model=MS-TCN2026.04 | 67.46 | 55.96 | 51.39 | 40.78 | 62.29 | — | |
| Timestamp Supervision approachFraction=0.012021.03 | 67.4 | — | — | — | — | — | |
| OriginalBackbone=MS-TCN2025.03 | 67.2 | 50.5 | 46.3 | 36.8 | 60.6 | — | |
| MS-TCNArchitecture Type=Standard, Base Backbone Model=MS-TCN2026.04 | 66.96 | 50.61 | 46.41 | 36.58 | 61.75 | — | |
| MS-TCNFeatures=I3D2019.03 | 66.3 | 52.6 | 48.1 | 37.9 | 61.7 | — | |
| MS-TCNBackbone=I3D2020.06 | 66.3 | 52.6 | 48.1 | 37.9 | 61.7 | — | |
| MSTCN2020.12 | 66.3 | 52.6 | 48.1 | 37.9 | 61.7 | — | |
| MS-TCNSupervision=Full2021.03 | 66.3 | 52.6 | 48.1 | 37.9 | 61.7 | — | |
| MS-TCNre-produced=false, refinement=false, grammar induction=None2023.12 | 66.3 | 52.6 | 48.1 | 37.9 | 61.7 | — | |
| MS-TCNre-prod.=false, refinement=false, grammar induction=none2023.12 | 66.3 | 52.6 | 48.1 | 37.9 | 61.7 | — | |
| MS-TCN2026.02 | 66.3 | 52.6 | 48.1 | 37.9 | 61.7 | — | |
| SSTDALabeled training data percentage=65%2020.03 | 65.8 | 69.3 | 62.9 | 49.4 | 69 | — | |
| Temporal AggregatesBackbone=I3D, Temporal scale=2s2020.06 | 65.3 | 52.3 | 46.5 | 34.8 | 51.3 | — | |
| MS-TCNFeatures=IDT2019.03 | 65.1 | 58.2 | 52.9 | 40.8 | 61.4 | — | |
| MS-TCNParams(M)=0.8, Backbone=I3D2024.09 | 65.1 | 58.2 | 52.9 | 40.8 | 61.4 | — | |
| MSTCN+GTRM2020.12 | 65 | 57.5 | 54 | 43.3 | 58.7 | — | |
| MS-TCNsource=official code2020.03 | 64.7 | 65.3 | 59.6 | 47.2 | 65.7 | — | |
| Temporal AggregatesBackbone=I3D, Temporal scale=5s2020.06 | 64.5 | 59.2 | 53.9 | 39.5 | 54.5 | — | |
| w/o constraintsSupervision Setting=Semi-supervised (5% labeled), Backbone=C2F-TCN2026.05 | 64.2 | 57.1 | 51.2 | 34.6 | 54.6 | — | |
| Timestamp Supervision approachSupervision=Timestamps2021.03 | 64.1 | 70.5 | 63.6 | 47.4 | 69.9 | — | |
| OursSupervision Setting=Semi-supervised (5% labeled), Backbone=C2F-TCN2026.05 | 63.9 | 58.5 | 52.4 | 36 | 56.4 | — | |
| Seg. Model M + plateauSupervision=Timestamps2021.03 | 63.5 | 65.5 | 59.1 | 43.2 | 65.9 | — | |
| Generative Network Inversion†Backbone=ASFormer2025.03 | 61.1 | 62.4 | 56 | 42.1 | 61.4 | — | |
| HMM-RNNFraction=0.12021.03 | 60.9 | — | — | — | — | — | |
| GRUFeatures=I3D2019.03 | 60.6 | — | — | — | — | — | |
| GRU2020.03 | 60.6 | — | — | — | — | — | |
| Generative Network InversionBackbone=ASFormer2025.03 | 59.8 | 54.1 | 47.7 | 34.1 | 48.8 | — | |
| HMM-RNNFraction=0.012021.03 | 58.8 | — | — | — | — | — | |
| OnlineTASInference mode=semi-online, Post-processing=false2024.11 | 57.4 | 17.8 | 14.8 | 10.1 | 19.6 | — | |
| OnlineTASInference mode=online, Post-processing=false2024.11 | 56.7 | 16.8 | 13.9 | 9.3 | 19.3 | — | |
| HTK(64)Features=IDT2019.03 | 56.3 | — | — | — | — | — | |
| Generative Network InversionBackbone=MS-TCN2025.03 | 55.5 | 46.7 | 41.1 | 28.7 | 45.6 | — | |
| Causal TCNInference mode=online2024.11 | 55.3 | 15.1 | 11.7 | 8.3 | 18.7 | — | |
| Generative Network Inversion†Backbone=MS-TCN2025.03 | 54.1 | 49.8 | 44.3 | 33.1 | 53.3 | — | |
| OnlineTASInference mode=semi-online, Post-processing=true2024.11 | 53.8 | 56.4 | 47.3 | 31.4 | 57.5 | — | |
| RoboSubtaskNet2026.02 | 53.5 | 48.4 | 41.7 | 30.4 | 52.03 | — | |
| OnlineTASInference mode=online, Post-processing=true2024.11 | 52.9 | 54.8 | 45.8 | 30.5 | 55.7 | — | |
| MeanBackbone=ASFormer2025.03 | 52.2 | 43.5 | 38.3 | 26.7 | 43.2 | — | |
| TCFPNFeatures=IDT2019.03 | 52 | — | — | — | — | — | |
| TCFPN2020.03 | 52 | — | — | — | — | — | |
| CoresetBackbone=ASFormer2025.03 | 52 | 48.3 | 42.4 | 29.7 | 48.1 | — | |
| HTKFeatures=IDT2019.03 | 50.7 | — | — | — | — | — | |
| CDFLSupervision=Transcripts2021.03 | 50.2 | — | — | — | — | — | |
| CoresetBackbone=MS-TCN2025.03 | 49.7 | 32.3 | 27.5 | 19.3 | 36.8 | — |