Research map / 3D vision / Depth and stereo
3D Crowd Counting via Geometric Attention-Guided Multi-view Fusion
Qi Zhang, Antoni B. Chan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace 2D average-height fusion (MVMS) with 3D projection and fusion | multi-view counting · PETS2009 | MAE: 3.49 -> 3.15 | Improved |
| replace MVMS with 3D counting plus geometric attention | multi-view counting · CityStreet | MAE: 8.01 -> 7.12 | Improved |
| add single-view 2D supervision branches to 3D loss (ablation) | 3D counting · PETS2009 (7*40cm) | MAE: 4.12 -> 3.20 | Improved |
| add projection consistency measure loss to 3D+2D (ablation) | 3D counting · DukeMTMC (18*20cm) | MAE: 1.63 -> 1.49 | Improved |
| use finer voxel height (10cm instead of 40cm) (ablation) | 3D counting · PETS2009 | MAE: 3.15 -> 4.21 | Hurt |
| use finer voxel height (10cm instead of 40cm) (ablation) | 3D counting · DukeMTMC | MAE: 1.65 -> 1.01 | Improved |
| add geometric attention along view ray (ablation) | 3D counting · DukeMTMC | MAE: 1.01 -> 0.92 | Improved |
| add geometric attention along view ray (ablation) | 3D counting · CityStreet | MAE: 7.54 -> 7.12 | Improved |
| add geometric attention along view ray (ablation) | 3D counting · PETS2009 | MAE: 3.15 -> 3.20 | Mixed |
| height map loss weight sigma too small or too large (ablation) | geometric attention · PETS2009 | MAE: 3.81 -> 3.15 -> 5.66 | Mixed |
| use cross entropy instead of MSE for height map loss (ablation) | geometric attention · PETS2009 | MAE: 3.15 -> 3.99 | Hurt |
| mask out background in height map loss (ablation) | geometric attention · CityStreet | MAE: 7.15 -> 7.72 | Hurt |
Closest papers by meaning
- 3D Crowd Counting via Multi-View Fusion with 3D Gaussian Kernels (AAAI 2020 · similarity 0.97)
- Wide-Area Crowd Counting via Ground-Plane Density Maps and Multi-View Fusion CNNs (CVPR 2019 · similarity 0.95)
- Wide-Area Crowd Counting: Multi-view Fusion Networks for Counting in Large Scenes (IJCV 2022 · similarity 0.95)
- Multi-Level Bottom-Top and Top-Bottom Feature Fusion for Crowd Counting (ICCV 2019 · similarity 0.91)
- Cross-View Cross-Scene Multi-View Crowd Counting (CVPR 2021 · similarity 0.90)
- Calibration-Free Multi-View Crowd Counting (ECCV 2022 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
