Self-supervised Spatio-temporal Representation Learning for Videos by Predicting Motion and Appearance Statistics

Benchmark Model Rank Results
self-supervised-action-recognition-on-hmdb51Motion & Appearance (C3D)#41Top-1 Accuracy: 20.3Pre-Training Dataset: UCF101Frozen: false
self-supervised-action-recognition-on-ucf101Motion & Appearance (C3D)#413-fold Accuracy: 58.8Pre-Training Dataset: UCF101Frozen: false