Cognitive Distraction Detection on DADA 2000 (Leave-one-dataset-out)
68.29AccuracyEyeCue
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| EyeCueBackbone=TimeSformerK600, Training-free=false2026.05 | 68.29 | 68 | |
| Voila-ABackbone=CLIP ViT-L/14, Training-free=false2026.05 | 67.68 | 67 | |
| Heatmap-basedBackbone=SVM, Training-free=false2026.05 | 65.85 | 64 | |
| TimeSformerK400Backbone=ViT-B/16, Training-free=false2026.05 | 62.63 | 63 | |
| EyeCueBackbone=VideoMAEK400, Training-free=false2026.05 | 62.5 | 62 | |
| TimeSformerK600Backbone=ViT-B/16, Training-free=false2026.05 | 61.89 | 53 | |
| VideoMAEK400Backbone=ViT-B/16, Training-free=false2026.05 | 61.28 | 60 | |
| GazeVQABackbone=CLIP ViT-B/32, Training-free=false2026.05 | 59.45 | 47 | |
| EgoVideoBackbone=ViT-B/14, Training-free=false2026.05 | 59.15 | 63 | |
| VideoLLaMA3Backbone=ViT-B/16, Training-free=false2026.05 | 53.26 | 38 | |
| Video-LLaVABackbone=OpenCLIP-L/14, Training-free=false2026.05 | 52.09 | 47 | |
| GazeGPTBackbone=GLM-4.5V-AWQ, Training-free=true2026.05 | 51.81 | 53 | |
| InternVideo2s1-1BBackbone=ViT-B/14, Training-free=false2026.05 | 50.51 | 45 | |
| GazeLLMBackbone=Gemini-2.5-Pro, Training-free=true2026.05 | 49.09 | 28 |