[
  {
    "Average accuracy of 3 splits": 85.1,
    "code_links": [
      {
        "title": "artest08/LateTemporalModeling3DCNN",
        "url": "https://github.com/artest08/LateTemporalModeling3DCNN"
      }
    ],
    "date": "2020-08-03",
    "date2": 20200803,
    "model": "R2+1D-BERT",
    "paper": {
      "title": "Late Temporal Modeling in 3D CNN Architectures with BERT for Action Recognition",
      "url": "https://cknow.io/lib/8266b47872f410ba"
    },
    "paper_data_uoa": "8266b47872f410ba"
  },
  {
    "Average accuracy of 3 splits": 82.48,
    "code_links": [],
    "date": "2019-06-13",
    "date2": 20190613,
    "model": "HAF+BoW/FV halluc",
    "paper": {
      "title": "Hallucinating IDT Descriptors and I3D Optical Flow Features for Action Recognition with CNNs",
      "url": "https://cknow.io/lib/45901c51dd6467aa"
    },
    "paper_data_uoa": "45901c51dd6467aa"
  },
  {
    "Average accuracy of 3 splits": 82.1,
    "code_links": [],
    "date": "2018-11-26",
    "date2": 20181126,
    "model": "EvaNet",
    "paper": {
      "title": "Evolving Space-Time Neural Architectures for Videos",
      "url": "https://cknow.io/lib/a44e4af3371c9b83"
    },
    "paper_data_uoa": "a44e4af3371c9b83"
  },
  {
    "Average accuracy of 3 splits": 81.9,
    "code_links": [],
    "date": "2019-08-27",
    "date2": 20190827,
    "model": "CCS + TSN (ImageNet+Kinetics pretrained)",
    "paper": {
      "title": "Cooperative Cross-Stream Network for Discriminative Action Representation",
      "url": "https://cknow.io/lib/09a74f888e9817da"
    },
    "paper_data_uoa": "09a74f888e9817da"
  },
  {
    "Average accuracy of 3 splits": 81.1,
    "code_links": [
      {
        "title": "piergiaj/representation-flow-cvpr19",
        "url": "https://github.com/piergiaj/representation-flow-cvpr19"
      }
    ],
    "date": "2018-10-02",
    "date2": 20181002,
    "model": "RepFlow-50",
    "paper": {
      "title": "Representation Flow for Action Recognition",
      "url": "https://cknow.io/lib/5354d908b72741e5"
    },
    "paper_data_uoa": "5354d908b72741e5"
  },
  {
    "Average accuracy of 3 splits": 80.92,
    "code_links": [],
    "date": "2019-03-20",
    "date2": 20190320,
    "model": "Multi-stream I3D ",
    "paper": {
      "title": "Contextual Action Cues from Camera Sensor for Multi-Stream Action Recognition",
      "url": "https://cknow.io/lib/e85e30290a1982be"
    },
    "paper_data_uoa": "e85e30290a1982be"
  },
  {
    "Average accuracy of 3 splits": 80.9,
    "code_links": [
      {
        "title": "craston/MARS",
        "url": "https://github.com/craston/MARS"
      }
    ],
    "date": "2019-06-01",
    "date2": 20190601,
    "model": "MARS+RGB+FLow (64 frames, Kinetics pretrained)",
    "paper": {
      "title": "MARS: Motion-Augmented RGB Stream for Action Recognition",
      "url": "https://cknow.io/lib/e917d94a27b04d45"
    },
    "paper_data_uoa": "e917d94a27b04d45"
  },
  {
    "Average accuracy of 3 splits": 80.9,
    "code_links": [
      {
        "title": "deepmind/kinetics-i3d",
        "url": "https://github.com/deepmind/kinetics-i3d"
      },
      {
        "title": "piergiaj/pytorch-i3d",
        "url": "https://github.com/piergiaj/pytorch-i3d"
      },
      {
        "title": "yaohungt/GSTEG_CVPR_2019",
        "url": "https://github.com/yaohungt/GSTEG_CVPR_2019"
      },
      {
        "title": "dlpbc/keras-kinetics-i3d",
        "url": "https://github.com/dlpbc/keras-kinetics-i3d"
      },
      {
        "title": "FrederikSchorr/sign-language",
        "url": "https://github.com/FrederikSchorr/sign-language"
      },
      {
        "title": "CMU-CREATE-Lab/deep-smoke-machine",
        "url": "https://github.com/CMU-CREATE-Lab/deep-smoke-machine"
      },
      {
        "title": "OanaIgnat/i3d_keras",
        "url": "https://github.com/OanaIgnat/i3d_keras"
      },
      {
        "title": "novav/CV_Spatiotemporal_Feature_3D_CNN",
        "url": "https://github.com/novav/CV_Spatiotemporal_Feature_3D_CNN"
      },
      {
        "title": "ahsaniqbal/Kinetics-FeatureExtractor",
        "url": "https://github.com/ahsaniqbal/Kinetics-FeatureExtractor"
      },
      {
        "title": "sebastiantiesmeyer/deeplabchop3d",
        "url": "https://github.com/sebastiantiesmeyer/deeplabchop3d"
      },
      {
        "title": "vijayvee/behavior_recognition",
        "url": "https://github.com/vijayvee/behavior_recognition"
      },
      {
        "title": "AbdurrahmanNadi/activity_recognition_web_service",
        "url": "https://github.com/AbdurrahmanNadi/activity_recognition_web_service"
      },
      {
        "title": "vijayvee/behavior-recognition",
        "url": "https://github.com/vijayvee/behavior-recognition"
      },
      {
        "title": "mHealthBuet/SegCodeNet",
        "url": "https://github.com/mHealthBuet/SegCodeNet"
      },
      {
        "title": "Alexyuda/action_recognition",
        "url": "https://github.com/Alexyuda/action_recognition"
      },
      {
        "title": "prinshul/GWSDR",
        "url": "https://github.com/prinshul/GWSDR"
      }
    ],
    "date": "2017-05-22",
    "date2": 20170522,
    "model": "Two-stream I3D",
    "paper": {
      "title": "Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset",
      "url": "https://cknow.io/lib/e93903d8f34491a9"
    },
    "paper_data_uoa": "e93903d8f34491a9"
  },
  {
    "Average accuracy of 3 splits": 80.5,
    "code_links": [
      {
        "title": "ZhaofanQiu/local-and-global-diffusion-networks",
        "url": "https://github.com/ZhaofanQiu/local-and-global-diffusion-networks"
      }
    ],
    "date": "2019-06-13",
    "date2": 20190613,
    "model": "LGD-3D Two-stream",
    "paper": {
      "title": "Learning Spatio-Temporal Representation with Local and Global Diffusion",
      "url": "https://cknow.io/lib/61784deb5c70a219"
    },
    "paper_data_uoa": "61784deb5c70a219"
  },
  {
    "Average accuracy of 3 splits": 78.7,
    "code_links": [
      {
        "title": "bryanyzhu/two-stream-pytorch",
        "url": "https://github.com/bryanyzhu/two-stream-pytorch"
      },
      {
        "title": "bryanyzhu/Hidden-Two-Stream",
        "url": "https://github.com/bryanyzhu/Hidden-Two-Stream"
      },
      {
        "title": "AbdalaDiasse/Video-classification-for-oil-quality-estimation",
        "url": "https://github.com/AbdalaDiasse/Video-classification-for-oil-quality-estimation"
      }
    ],
    "date": "2017-04-02",
    "date2": 20170402,
    "model": "Hidden Two-Stream",
    "paper": {
      "title": "Hidden Two-Stream Convolutional Networks for Action Recognition",
      "url": "https://cknow.io/lib/1b3bdadb4a42b57d"
    },
    "paper_data_uoa": "1b3bdadb4a42b57d"
  },
  {
    "Average accuracy of 3 splits": 76.5,
    "code_links": [
      {
        "title": "holistic-video-understanding/HVU-Dataset",
        "url": "https://github.com/holistic-video-understanding/HVU-Dataset"
      },
      {
        "title": "holistic-video-understanding/Mini-HVU",
        "url": "https://github.com/holistic-video-understanding/Mini-HVU"
      }
    ],
    "date": "2019-04-25",
    "date2": 20190425,
    "model": "HATNet (32 frames)",
    "paper": {
      "title": "Large Scale Holistic Video Understanding",
      "url": "https://cknow.io/lib/5b16b09584bdea1a"
    },
    "paper_data_uoa": "5b16b09584bdea1a"
  },
  {
    "Average accuracy of 3 splits": 75.7,
    "code_links": [],
    "date": "2019-06-10",
    "date2": 20190610,
    "model": "FASTER32 (Kinetics pretrain)",
    "paper": {
      "title": "FASTER Recurrent Networks for Efficient Video Classification",
      "url": "https://cknow.io/lib/474ff5967144a33a"
    },
    "paper_data_uoa": "474ff5967144a33a"
  },
  {
    "Average accuracy of 3 splits": 74.3,
    "code_links": [],
    "date": "2018-07-24",
    "date2": 20180724,
    "model": "ADL+ResNet+IDT",
    "paper": {
      "title": "Contrastive Video Representation Learning via Adversarial Perturbations",
      "url": "https://cknow.io/lib/1e9f286c44f46a11"
    },
    "paper_data_uoa": "1e9f286c44f46a11"
  },
  {
    "Average accuracy of 3 splits": 74.2,
    "code_links": [
      {
        "title": "kevin-ssy/Optical-Flow-Guided-Feature",
        "url": "https://github.com/kevin-ssy/Optical-Flow-Guided-Feature"
      }
    ],
    "date": "2017-11-29",
    "date2": 20171129,
    "model": "Optical Flow Guided Feature",
    "paper": {
      "title": "Optical Flow Guided Feature: A Fast and Robust Motion Representation for Video Action Recognition",
      "url": "https://cknow.io/lib/47c4aeda535c5d28"
    },
    "paper_data_uoa": "47c4aeda535c5d28"
  },
  {
    "Average accuracy of 3 splits": 72.2,
    "code_links": [
      {
        "title": "feichtenhofer/st-resnet",
        "url": "https://github.com/feichtenhofer/st-resnet"
      }
    ],
    "date": "2017-07-01",
    "date2": 20170701,
    "model": "STM Network+IDT",
    "paper": {
      "title": "Spatiotemporal Multiplier Networks for Video Action Recognition",
      "url": "https://cknow.io/lib/002d244a408c4574"
    },
    "paper_data_uoa": "002d244a408c4574"
  },
  {
    "Average accuracy of 3 splits": 72.2,
    "code_links": [],
    "date": "2019-08-07",
    "date2": 20190807,
    "model": "STM (ImageNet+Kinetics pretrain)",
    "paper": {
      "title": "STM: SpatioTemporal and Motion Encoding for Action Recognition",
      "url": "https://cknow.io/lib/dbf65072610598a3"
    },
    "paper_data_uoa": "dbf65072610598a3"
  },
  {
    "Average accuracy of 3 splits": 72,
    "code_links": [],
    "date": "2019-04-05",
    "date2": 20190405,
    "model": "Prob-Distill",
    "paper": {
      "title": "Attention Distillation for Learning Video Representations",
      "url": "https://cknow.io/lib/4d88a88b51e44754"
    },
    "paper_data_uoa": "4d88a88b51e44754"
  },
  {
    "Average accuracy of 3 splits": 71.5,
    "code_links": [],
    "date": "2020-02-11",
    "date2": 20200211,
    "model": "TesNet (ImageNet pretrained)",
    "paper": {
      "title": "Learning spatio-temporal representations with temporal squeeze pooling",
      "url": "https://cknow.io/lib/ab18c88ae422e84b"
    },
    "paper_data_uoa": "ab18c88ae422e84b"
  },
  {
    "Average accuracy of 3 splits": 71.13,
    "code_links": [],
    "date": "2019-05-29",
    "date2": 20190529,
    "model": "HF-ECOLite (ImageNet+Kinetics pretrain)",
    "paper": {
      "title": "Hierarchical Feature Aggregation Networks for Video Action Recognition",
      "url": "https://cknow.io/lib/dd1247b8980e18a8"
    },
    "paper_data_uoa": "dd1247b8980e18a8"
  },
  {
    "Average accuracy of 3 splits": 69,
    "code_links": [
      {
        "title": "jeffreyyihuang/two-stream-action-recognition",
        "url": "https://github.com/jeffreyyihuang/two-stream-action-recognition"
      },
      {
        "title": "jeffreyhuang1/two-stream-action-recognition",
        "url": "https://github.com/jeffreyhuang1/two-stream-action-recognition"
      },
      {
        "title": "chihyaoma/Activity-Recognition-with-CNN-and-RNN",
        "url": "https://github.com/chihyaoma/Activity-Recognition-with-CNN-and-RNN"
      }
    ],
    "date": "2017-03-30",
    "date2": 20170330,
    "model": "TS-LSTM",
    "paper": {
      "title": "TS-LSTM and Temporal-Inception: Exploiting Spatiotemporal Dynamics for Activity Recognition",
      "url": "https://cknow.io/lib/471ea07a997aee31"
    },
    "paper_data_uoa": "471ea07a997aee31"
  },
  {
    "Average accuracy of 3 splits": 68.4,
    "code_links": [
      {
        "title": "piergiaj/latent-subevents",
        "url": "https://github.com/piergiaj/latent-subevents"
      }
    ],
    "date": "2016-05-26",
    "date2": 20160526,
    "model": "Sub-events",
    "paper": {
      "title": "Learning Latent Sub-events in Activity Videos Using Temporal Attention Filters",
      "url": "https://cknow.io/lib/256bbd4333843d3a"
    },
    "paper_data_uoa": "256bbd4333843d3a"
  },
  {
    "Average accuracy of 3 splits": 65.4,
    "code_links": [
      {
        "title": "feichtenhofer/twostreamfusion",
        "url": "https://github.com/feichtenhofer/twostreamfusion"
      }
    ],
    "date": "2016-04-22",
    "date2": 20160422,
    "model": "S:VGG-16, T:VGG-16",
    "paper": {
      "title": "Convolutional Two-Stream Network Fusion for Video Action Recognition",
      "url": "https://cknow.io/lib/6221d483c1e04712"
    },
    "paper_data_uoa": "6221d483c1e04712"
  },
  {
    "Average accuracy of 3 splits": 64.8,
    "code_links": [
      {
        "title": "gulvarol/ltc",
        "url": "https://github.com/gulvarol/ltc"
      }
    ],
    "date": "2016-04-15",
    "date2": 20160415,
    "model": "LTC",
    "paper": {
      "title": "Long-term Temporal Convolutions for Action Recognition",
      "url": "https://cknow.io/lib/656933d442a7985c"
    },
    "paper_data_uoa": "656933d442a7985c"
  },
  {
    "Average accuracy of 3 splits": 62.8,
    "code_links": [],
    "date": "2019-06-19",
    "date2": 20190619,
    "model": "R-STAN-50",
    "paper": {
      "title": "R-STAN: Residual Spatial-Temporal Attention Network for Action Recognition",
      "url": "https://cknow.io/lib/1c94b929dfcdb35e"
    },
    "paper_data_uoa": "1c94b929dfcdb35e"
  },
  {
    "Average accuracy of 3 splits": 62.7,
    "code_links": [],
    "date": "2018-12-03",
    "date2": 20181203,
    "model": "SUSiNet (multi, Kinetics pretrained)",
    "paper": {
      "title": "SUSiNet: See, Understand and Summarize it",
      "url": "https://cknow.io/lib/2c70a2a80d451699"
    },
    "paper_data_uoa": "2c70a2a80d451699"
  },
  {
    "Average accuracy of 3 splits": 59.4,
    "code_links": [
      {
        "title": "woodfrog/ActionRecognition",
        "url": "https://github.com/woodfrog/ActionRecognition"
      },
      {
        "title": "damien911224/theWorldInSafety",
        "url": "https://github.com/damien911224/theWorldInSafety"
      },
      {
        "title": "mcgridles/LENS",
        "url": "https://github.com/mcgridles/LENS"
      },
      {
        "title": "jerryljq/ActionRecognition",
        "url": "https://github.com/jerryljq/ActionRecognition"
      }
    ],
    "date": "2014-06-09",
    "date2": 20140609,
    "model": "Two-stream",
    "paper": {
      "title": "Two-Stream Convolutional Networks for Action Recognition in Videos",
      "url": "https://cknow.io/lib/3b055425da7fcf4d"
    },
    "paper_data_uoa": "3b055425da7fcf4d"
  },
  {
    "Average accuracy of 3 splits": 55.16,
    "code_links": [],
    "date": "2019-06-19",
    "date2": 20190619,
    "model": "R-STAN-152",
    "paper": {
      "title": "R-STAN: Residual Spatial-Temporal Attention Network for Action Recognition",
      "url": "https://cknow.io/lib/1c94b929dfcdb35e"
    },
    "paper_data_uoa": "1c94b929dfcdb35e"
  },
  {
    "Average accuracy of 3 splits": 54.8,
    "code_links": [],
    "date": "2019-01-26",
    "date2": 20190126,
    "model": "R(2+1)D-18 (DistInit pretraining)",
    "paper": {
      "title": "DistInit: Learning Video Representations Without a Single Labeled Video",
      "url": "https://cknow.io/lib/a410b42da91f2d04"
    },
    "paper_data_uoa": "a410b42da91f2d04"
  },
  {
    "Average accuracy of 3 splits": 51.8,
    "code_links": [],
    "date": "2018-03-22",
    "date2": 20180322,
    "model": "CD-UAR",
    "paper": {
      "title": "Towards Universal Representation for Unseen Action Recognition",
      "url": "https://cknow.io/lib/9c5347f4953a03dc"
    },
    "paper_data_uoa": "9c5347f4953a03dc"
  },
  {
    "Average accuracy of 3 splits": 33.4,
    "code_links": [
      {
        "title": "laura-wang/video_repres_mas",
        "url": "https://github.com/laura-wang/video_repres_mas"
      }
    ],
    "date": "2019-04-07",
    "date2": 20190407,
    "model": "Pretrained on Kinetics",
    "paper": {
      "title": "Self-supervised Spatio-temporal Representation Learning for Videos by Predicting Motion and Appearance Statistics",
      "url": "https://cknow.io/lib/510450184304f519"
    },
    "paper_data_uoa": "510450184304f519"
  },
  {
    "code_links": [],
    "date": "2020-06-05",
    "date2": 20200605,
    "model": "GDT",
    "paper": {
      "title": "Multi-modal Self-Supervision from Generalized Data Transformations",
      "url": "https://cknow.io/lib/6cc21bd044a7838e"
    },
    "paper_data_uoa": "6cc21bd044a7838e"
  }
]