Research map / Detection and segmentation / 3D detection and driving
Learning Pseudo 3D Representation for Ego-centric 2D Multiple Object Tracking
Jiawei He, Lue Fan, Zhaoxiang Zhang
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 low-quality detection matching (ablation) | modified DeepSORT baseline · WOD val FRONT camera | MOTA: 51.0 -> 53.4 | Improved |
| add GNN association learning (ablation) | DeepSORT + low-quality dets · WOD val FRONT camera | MOTA: 53.4 -> 56.5 | Improved |
| add jointly learned 3D representation (ablation) | P3DTrack · WOD val FRONT camera | MOTA: 56.5 -> 57.6 | Improved |
| add jointly learned 3D representation (ablation) | P3DTrack · WOD val FRONT camera | IDF1: 66.5 -> 68.1 | Improved |
| add 2D Kalman motion to 2D feature (ablation) | P3DTrack · WOD val FRONT camera | IDF1: 66.5 -> 58.6 | Hurt |
| use 3D representation instead of 3D Kalman motion without learned representation (ablation) | P3DTrack · WOD val FRONT camera | IDF1: 63.3 -> 68.1 | Improved |
| use learned 3D representation instead of SfM 3D positions (ablation) | P3DTrack · WOD val FRONT camera | MOTA: 55.5 -> 57.6 | Improved |
| use learned 3D representation instead of MiDaS v3 depth (ablation) | P3DTrack · WOD val FRONT camera | IDF1: 63.4 -> 68.1 | Improved |
| raise detection threshold from 0.5 to 0.6 (ablation) | P3DTrack · WOD val FRONT camera | MOTA: 57.6 -> 50.1 | Hurt |
| lower detection threshold from 0.5 to 0.4 (ablation) | P3DTrack · WOD val FRONT camera | MOTA: 57.6 -> 51.4 | Hurt |
| raise appearance threshold from 0.6 to 0.7 (ablation) | P3DTrack · WOD val FRONT camera | MOTA: 57.6 -> 58.0 | Mixed |
| use 3D similarity only (alpha 1.0) instead of alpha 0.4 (ablation) | P3DTrack · WOD val FRONT camera | IDF1: 68.1 -> 54.2 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- QDTrack — QDTrack: Quasi-Dense Similarity Learning for Appearance-Only Multiple Object Tracking (TPAMI 2023) · better on 2 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
- 3DMOTFormer: Graph Transformer for Online 3D Multi-Object Tracking (ICCV 2023 · similarity 0.89)
- GNN3DMOT: Graph Neural Network for 3D Multi-Object Tracking With 2D-3D Multi-Feature Learning (CVPR 2020 · similarity 0.89)
- S2-Track: A Simple yet Strong Approach for End-to-End 3D Multi-Object Tracking (ICML 2025 · similarity 0.88)
- GRAE-3DMOT: Geometry Relation-Aware Encoder for Online 3D Multi-Object Tracking (CVPR 2025 · similarity 0.87)
- ADA-Track: End-to-End Multi-Camera 3D Multi-Object Tracking with Alternating Detection and Association (CVPR 2024 · similarity 0.87)
- M3SOT: Multi-Frame, Multi-Field, Multi-Space 3D Single Object Tracking (AAAI 2024 · similarity 0.86)
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.
