Human attention prediction on BDD-A DKL > 2 non-trivial frames (test)
1.67Mean KL DivergenceBDD-A prediction model (HWS)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| BDD-A prediction model (HWS)Sampling Strategy=Human Weighted Sampling (HWS), Output Size=64x362017.11 | 1.67 | — | 0.44 | — | |
| BDD-A prediction model (default)Sampling Strategy=default, Output Size=64x362017.11 | 1.71 | — | 0.41 | — | |
| SALICONFine-tuned=true, Implementation=Open source [26], Output Size=64x36 (scaled)2017.11 | 1.76 | — | 0.39 | — | |
| Averaged attention baselineDescription=Predicts the averaged human attention map of training videos2017.11 | 1.87 | — | 0.36 | — | |
| DR(eye)VEPre-trained=true, Fine-tuned=false, Output Size=64x36 (scaled)2017.11 | 2.63 | — | 0.35 | — |