Research map / 3D vision / Face and shape
Geometry-Constrained Car Recognition Using a 3D Perspective Network
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 attention-based feature (RoI Align) with 3D perspective feature (ablation) | Ours-det · BoxCars Medium | top-1 accuracy: 73.97 -> 78.45 | Helped |
| replace attention-based feature with 3D perspective feature (ablation) | Ours-det · BoxCars Hard | top-1 accuracy: 73.65 -> 75.18 | Helped |
| replace ABF with 3DPF (ablation) | Ours-det · BoxCars Hard | top-5 accuracy: 91.13 -> 91.53 | No clear effect |
| add VP regularization loss (ablation) | Ours-det · BoxCars Medium | top-1 accuracy: 78.32 -> 78.45 | No clear effect |
| add VP regularization loss (ablation) | Ours-det · BoxCars Hard | top-5 accuracy: 89.94 -> 91.53 | Helped |
| add VP regularization loss (ablation) | Ours-cls · BoxCars Medium | top-5 accuracy: 92.88 -> 93.82 | Helped |
| add VP regularization loss (ablation) | Ours-cls · BoxCars Hard | top-5 accuracy: 90.50 -> 91.97 | Helped |
| add VP regularization loss (ablation) | Ours-cls · BoxCars Hard | top-1 accuracy: 76.84 -> 77.08 | No clear effect |
| add VP loss, 3D box PCK at 15 epochs (ablation) | Ours-det · BoxCars Hard | PCK: 86.70 -> 87.15 | Helped |
| add VP loss, cuboid quality at 15 epochs (ablation) | Ours-det · BoxCars Hard | CQ: 1.69 -> 2.12 | Helped |
| add VP loss, cuboid quality at 5 epochs (ablation) | Ours-det · BoxCars Hard | CQ: 1.48 -> 1.48 | No clear effect |
| add 3D perspective features to 2D detector | Ours-det vs Faster-RCNN · BoxCars Hard | top-1 accuracy: 62.73 -> 75.18 | Helped |
Closest papers by meaning
- PerspectiveNet: 3D Object Detection from a Single RGB Image via Perspective Points (NeurIPS 2019 · similarity 0.88)
- BoxCars: 3D Boxes as CNN Input for Improved Fine-Grained Vehicle Recognition (CVPR 2016 · similarity 0.87)
- Exploring intermediate representation for monocular vehicle pose estimation (CVPR 2021 · similarity 0.87)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 2024 · similarity 0.84)
- Infer the Whole from a Glimpse of a Part: Keypoint-Based Knowledge Graph for Vehicle Re-Identification (AAAI 2025 · similarity 0.83)
- Tri-Perspective View Decomposition for Geometry-Aware Depth Completion (CVPR 2024 · similarity 0.83)
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.