Research map / 3D vision / Human pose estimation
GSNet: Joint Vehicle Pose and Shape Reconstruction with Geometrical and Scene-aware Supervision
Lei Ke, Shichao Li, Yanan Sun, Yu-Wing Tai, Chi-Keung Tang
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 |
|---|---|---|---|
| add detected box features to ROI features (ablation) | GSNet · ApolloCar3D (validation) | A3DP-Rel mean: 6.8 -> 12.5 | Improved |
| add heatmap features (ablation) | GSNet · ApolloCar3D (validation) | A3DP-Rel mean: 12.5 -> 13.7 | Improved |
| add keypoint position features (ablation) | GSNet · ApolloCar3D (validation) | A3DP-Rel mean: 13.7 -> 14.1 | Improved |
| add box features (ablation) | GSNet · ApolloCar3D (validation) | translation error T: 2.41 -> 1.56 | Improved |
| add 3D shape reconstruction loss (ablation) | GSNet · ApolloCar3D (validation) | A3DP-Rel mean: 14.1 -> 15.1 | Improved |
| add geometric keypoint reprojection loss (ablation) | GSNet · ApolloCar3D (validation) | A3DP-Rel mean: 15.1 -> 17.6 | Improved |
| add inter-instance scene-aware loss (ablation) | GSNet · ApolloCar3D (validation) | A3DP-Rel c-s: 14.2 -> 16.3 | Improved |
| add intra-instance co-planar loss (ablation) | GSNet · ApolloCar3D (validation) | A3DP-Rel mean: 18.8 -> 20.2 | Improved |
| use divide-and-conquer shape module instead of single PCA (ablation) | GSNet · ApolloCar3D (validation) | shape reconstruction error: 88.68 -> 81.33 | Improved |
| use divide-and-conquer shape module instead of retrieval (ablation) | GSNet · ApolloCar3D (validation) | A3DP-Rel mean: 17.6 -> 20.2 | Improved |
| use GSNet instead of 3D-RCNN | GSNet · ApolloCar3D (test) | A3DP-Rel mean: 10.79 -> 20.21 | Improved |
| use GSNet instead of 3D-RCNN | GSNet · Pascal3D+ car viewpoint | MedErr: 3.0 -> 2.4 | Improved |
Later papers that compared with this method
6 later accepted papers report GSNet as a baseline in their results tables; 6 of them report a better number on most of the results they share with it.
- GraphGrasp: Lightweight and Efficient Graph-Guided 6-DoF Robotic Grasp Pose Estimation Network (AAAI 2026 · better on 4 of 5 shared results)
- A Cross-view Fusion Framework for Robust 6-DoF Grasp Pose Estimation (CVPR 2026 · better on 4 of 4 shared results)
- Fixed Reality, Diffused Possibility: Disentangling Stochastic and Deterministic Latent for Cluttered Grasping (ECCV 2026 · better on 3 of 3 shared results)
- Monocular Vehicle Pose and Shape Reconstruction via Dynamic Context Adaptation and Progressive Geometry Refinement (AAAI 2026 · better on 2 of 2 shared results)
- Local Occupancy-Enhanced Object Grasping with Multiple Triplanar Projection (ECCV 2024 · better on 5 of 5 shared results)
- BAAM: Monocular 3D Pose and Shape Reconstruction With Bi-Contextual Attention Module and Attention-Guided Modeling (CVPR 2023 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DeepMANTA — Deep MANTA: A Coarse-To-Fine Many-Task Network for Joint 2D and 3D Vehicle Analysis From Monocular Image (CVPR 2017) · better on 3 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- GSIR: Generalizable 3D Shape Interpretation and Reconstruction (ECCV 2020 · similarity 0.86)
- Part-level Car Parsing and Reconstruction in Single Street View Images (TPAMI 2021 · similarity 0.86)
- GCE-Pose: Global Context Enhancement for Category-level Object Pose Estimation (CVPR 2025 · similarity 0.85)
- ViiNeuS: Volumetric Initialization for Implicit Neural Surface Reconstruction of Urban Scenes with Limited Image Overlap (CVPR 2025 · similarity 0.85)
- SS3DM: Benchmarking Street-View Surface Reconstruction with a Synthetic 3D Mesh Dataset (NeurIPS 2024 · similarity 0.84)
- Learning Generalizable Shape Completion with SIM(3) Equivariance (NeurIPS 2025 · similarity 0.84)
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.
