Group Activity Detection on Cafe Split-by-Place
17.03Group mAP@1.0ProGraD (Decoder + Prompts)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ProGraD (Decoder + Prompts)Backbone=DINOv2, Full FT=false, Trainable Param(M)=10.742025.08 | 17.03 | 39.11 | 67.89 | |
| ProGraD (Decoder + Adapter)Backbone=DINOv2, Full FT=false, Trainable Param(M)=11.872025.08 | 16.18 | 33.4 | 66.94 | |
| ProGraD (Decoder Only)Backbone=DINOv2, Full FT=false, Trainable Param(M)=10.682025.08 | 13.02 | 32.68 | 65.44 | |
| VLCMTNBackbone=SAM, Full FT=false, Trainable Param(M)=∼ 24*2025.08 | 11.58 | 30.98 | 65.48 | |
| Practical GADBackbone=ResNet-18, Full FT=true, Trainable Param(M)=22.772025.08 | 10.85 | 30.9 | 63.84 | |
| JRDB-baseBackbone=I3D, Full FT=true, Trainable Param(M)=∼16*2025.08 | 9.42 | 25.75 | 48 | |
| JointBackbone=I3D, Full FT=true, Trainable Param(M)=∼17*2025.08 | 8.39 | 26.26 | 55.95 | |
| ARGBackbone=Inception-v3, Full FT=true, Trainable Param(M)=∼52*2025.08 | 6.87 | 28.44 | 46.72 | |
| HGCBackbone=I3D, Full FT=true, Trainable Param(M)=∼20*2025.08 | 3.47 | 18.46 | 52.56 |