# Action Recognition on UCF-101 (3-Fold Cross-Validation) Source: [Action Recognition on UCF-101 (3-Fold Cross-Validation)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:21:11.036Z Evaluation of video action recognition performance using the UCF-101 dataset, typically measured using 3-fold cross-validation accuracy. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Action Recognition | | Dataset (including any stated split) | UCF-101 | | Metric | 3-Fold Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 32 of 32 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [3-Fold Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | 3-Fold Accuracy | Result date | Report paper | | --- | --- | --- | --- | | OmniVec | 99.6 | Not recorded | 68fa7a181e209131cea811ca | | OmniVec2 | 99.1 | Not recorded | 68fa7a181e209131cea811ca | | CoViAR + flow | 94.9 | Not recorded | 68fa64301e209131cea03528 | | Very deep two-stream | 91.4 | Not recorded | 68fa61ff1e209131ce9f70ca | | CoViAR | 90.4 | Not recorded | 68fa64301e209131cea03528 | | TDD+FV | 90.3 | Not recorded | 68fa61ff1e209131ce9f70ca | | MIFS+FV | 89.1 | Not recorded | 68fa61ff1e209131ce9f70ca | | Flow | 89 | Not recorded | 68fa64301e209131cea03528 | | LSTM | 88.6 | Not recorded | 68fa61d01e209131ce9f634a | | Two-stream+LSTM | 88.6 | Not recorded | 68fa61ff1e209131ce9f70ca | | Conv Pooling | 88.2 | Not recorded | 68fa61d01e209131ce9f634a | | Two-stream model | 88 | Not recorded | 68fa61951e209131ce9f4ccf | | Two-Stream CNN | 88 | Not recorded | 68fa61d01e209131ce9f634a | | Two-stream | 88 | Not recorded | 68fa61ff1e209131ce9f70ca | | IDT with higher-dimensional encodings | 87.9 | Not recorded | 68fa61951e209131ce9f4ccf | | Improved Dense Trajectories (IDTF)s | 87.9 | Not recorded | 68fa61d01e209131ce9f634a | | iDT+HSV | 87.9 | Not recorded | 68fa61ff1e209131ce9f70ca | | Conv Pooling | 87.6 | Not recorded | 68fa61d01e209131ce9f634a | | Two-stream model | 86.9 | Not recorded | 68fa61951e209131ce9f4ccf | | Two-Stream CNN | 86.9 | Not recorded | 68fa61d01e209131ce9f634a | | Improved dense trajectories (IDT) | 85.9 | Not recorded | 68fa61951e209131ce9f4ccf | | iDT+FV | 85.9 | Not recorded | 68fa61ff1e209131ce9f70ca | | Temporal stream ConvNet | 83.7 | Not recorded | 68fa61951e209131ce9f4ccf | | Single Frame CNN Model | 73.9 | Not recorded | 68fa61d01e209131ce9f634a | | Single Frame Model | 73.3 | Not recorded | 68fa61d01e209131ce9f634a | | Spatial stream ConvNet | 73 | Not recorded | 68fa61951e209131ce9f4ccf | | Single Frame CNN Model | 73 | Not recorded | 68fa61d01e209131ce9f634a | | Slow fusion spatio-temporal ConvNet | 65.4 | Not recorded | 68fa61951e209131ce9f4ccf | | Slow Fusion CNN | 65.4 | Not recorded | 68fa61d01e209131ce9f634a | | DeepNet | 63.3 | Not recorded | 68fa61ff1e209131ce9f70ca | | Tuple verification | 50.2 | Not recorded | 68fa62a31e209131ce9fa753 | | Random | 38.6 | Not recorded | 68fa62a31e209131ce9fa753 | ### OmniVec 3-Fold Accuracy: 99.6. Result ID: 698b6d317a967eddb76e7847. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Modality":"Video"}. This record does not assert independent reproduction. ### OmniVec2 3-Fold Accuracy: 99.1. Result ID: 698b6d317a967eddb76e7846. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Modality":"Video"}. This record does not assert independent reproduction. ### CoViAR + flow 3-Fold Accuracy: 94.9. Result ID: 698c7f3e37894ef7d9ced012. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Input Streams":"Compressed Representation + Optical Flow"}. This record does not assert independent reproduction. ### Very deep two-stream 3-Fold Accuracy: 91.4. Result ID: 698c672b37894ef7d9ce2275. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Year":"2015"}. This record does not assert independent reproduction. ### CoViAR 3-Fold Accuracy: 90.4. Result ID: 698c7f3e37894ef7d9ced010. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Input Streams":"Compressed Representation"}. This record does not assert independent reproduction. ### TDD+FV 3-Fold Accuracy: 90.3. Result ID: 698c672b37894ef7d9ce2271. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Year":"2015"}. This record does not assert independent reproduction. ### MIFS+FV 3-Fold Accuracy: 89.1. Result ID: 698c672b37894ef7d9ce2270. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Year":"2015"}. This record does not assert independent reproduction. ### Flow 3-Fold Accuracy: 89. Result ID: 698c7f3e37894ef7d9ced011. [Reporting paper]() Recorded evaluation configuration: {"Input Streams":"Optical Flow"}. This record does not assert independent reproduction. ### LSTM 3-Fold Accuracy: 88.6. Result ID: 698c669d37894ef7d9ce1f6b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inputs":"Optical Flow + Image Frames","Unroll Frames":30}. This record does not assert independent reproduction. ### Two-stream+LSTM 3-Fold Accuracy: 88.6. Result ID: 698c672b37894ef7d9ce2274. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Year":"2015"}. This record does not assert independent reproduction. ### Conv Pooling 3-Fold Accuracy: 88.2. Result ID: 698c669d37894ef7d9ce1f6a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inputs":"Image Frames + Optical Flow","Frames":120}. This record does not assert independent reproduction. ### Two-stream model 3-Fold Accuracy: 88. Result ID: 698c661437894ef7d9ce1878. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Fusion Method":"SVM"}. This record does not assert independent reproduction. ### Two-Stream CNN 3-Fold Accuracy: 88. Result ID: 698c669d37894ef7d9ce1f67. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inputs":"Optical Flow + Image Frames","Fusion Method":"SVM Fusion"}. This record does not assert independent reproduction. ### Two-stream 3-Fold Accuracy: 88. Result ID: 698c672b37894ef7d9ce2273. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Year":"2014"}. This record does not assert independent reproduction. ### IDT with higher-dimensional encodings 3-Fold Accuracy: 87.9. Result ID: 698c661437894ef7d9ce1873. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Improved Dense Trajectories (IDTF)s 3-Fold Accuracy: 87.9. Result ID: 698c669d37894ef7d9ce1f62. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### iDT+HSV 3-Fold Accuracy: 87.9. Result ID: 698c672b37894ef7d9ce226f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Year":"2014"}. This record does not assert independent reproduction. ### Conv Pooling 3-Fold Accuracy: 87.6. Result ID: 698c669d37894ef7d9ce1f69. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inputs":"Image Frames + Optical Flow","Frames":30}. This record does not assert independent reproduction. ### Two-stream model 3-Fold Accuracy: 86.9. Result ID: 698c661437894ef7d9ce1877. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Fusion Method":"averaging"}. This record does not assert independent reproduction. ### Two-Stream CNN 3-Fold Accuracy: 86.9. Result ID: 698c669d37894ef7d9ce1f66. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inputs":"Optical Flow + Image Frames","Fusion Method":"Averaging"}. This record does not assert independent reproduction. ### Improved dense trajectories (IDT) 3-Fold Accuracy: 85.9. Result ID: 698c661437894ef7d9ce1872. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### iDT+FV 3-Fold Accuracy: 85.9. Result ID: 698c672b37894ef7d9ce226e. [Reporting paper]() Recorded evaluation configuration: {"Year":"2013"}. This record does not assert independent reproduction. ### Temporal stream ConvNet 3-Fold Accuracy: 83.7. Result ID: 698c661437894ef7d9ce1876. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Stream":"Temporal"}. This record does not assert independent reproduction. ### Single Frame CNN Model 3-Fold Accuracy: 73.9. Result ID: 698c669d37894ef7d9ce1f65. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inputs":"Optical Flow"}. This record does not assert independent reproduction. ### Single Frame Model 3-Fold Accuracy: 73.3. Result ID: 698c669d37894ef7d9ce1f68. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Spatial stream ConvNet 3-Fold Accuracy: 73. Result ID: 698c661437894ef7d9ce1875. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Stream":"Spatial"}. This record does not assert independent reproduction. ### Single Frame CNN Model 3-Fold Accuracy: 73. Result ID: 698c669d37894ef7d9ce1f64. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inputs":"Images"}. This record does not assert independent reproduction. ### Slow fusion spatio-temporal ConvNet 3-Fold Accuracy: 65.4. Result ID: 698c661437894ef7d9ce1874. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Slow Fusion CNN 3-Fold Accuracy: 65.4. Result ID: 698c669d37894ef7d9ce1f63. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### DeepNet 3-Fold Accuracy: 63.3. Result ID: 698c672b37894ef7d9ce2272. [Reporting paper]() Recorded evaluation configuration: {"Year":"2014"}. This record does not assert independent reproduction. ### Tuple verification 3-Fold Accuracy: 50.2. Result ID: 698c6bcd37894ef7d9ce4110. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Initialization":"Tuple verification"}. This record does not assert independent reproduction. ### Random 3-Fold Accuracy: 38.6. Result ID: 698c6bcd37894ef7d9ce410f. [Reporting paper]() Recorded evaluation configuration: {"Initialization":"Random"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.