Research map / Detection and segmentation / 3D detection and driving
PLOT: Pseudo-Labeling via Object Tracking for Monocular 3D Object Detection
SeokYeong Lee, Sithu Aung, JunYong Choi, Seungryong Kim, Ig-Jae Kim, Junghyun Cho
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 |
|---|---|---|---|
| fuse more adjacent frames (1 to 20) in trajectory-guided shape fusion (ablation) | PLOT · KITTI Car val | AP3D@0.3 Easy: 51.75 -> 80.48 | Improved |
| increase fusion window from +8 to +20 frames (ablation) | PLOT · KITTI Car val | AP3D@0.3 Easy: 77.23 -> 80.48 | Improved |
| train MonoDETR on pseudo-labels instead of using raw labels (ablation) | PLOT · KITTI Car val | AP3D@0.3 Easy: 59.82 -> 80.48 | Improved |
| add global object memory (ablation) | PLOT · KITTI-360 Car | AP3D@0.3 Easy: 48.14 -> 54.78 | Improved |
| add global object memory (ablation) | PLOT · KITTI-360 Car | AP3D@0.3 Hard: 40.28 -> 48.75 | Improved |
| add camera-motion-based static/dynamic reasoning for orientation (ablation) | PLOT · KITTI Car val | AP3D@0.3 Easy: 77.65 -> 80.48 | Improved |
| use ground-truth intrinsics instead of UniDepth-estimated (ablation) | PLOT · KITTI Car val | AP3D@0.3 Easy: 70.44 -> 80.48 | Improved |
| use the accurate AllTracker instead of a tiny tracker (ablation) | PLOT · Waymo scene | AP3D@0.3 Easy: 20.57 -> 20.89 | No clear effect |
| inject temporal scale jitter 0.04 into depth (ablation) | PLOT (SE(3)) · Waymo scene | AP3D@0.3 Easy: 20.89 -> 14.14 | Hurt |
| inject range noise 0.08 into depth (ablation) | PLOT (SE(3)) · Waymo scene | AP3D@0.3 Easy: 20.89 -> 19.48 | Hurt |
| use Sim(3) alignment with scale term under temporal jitter 0.04 (ablation) | PLOT · Waymo scene | AP3D@0.3 Easy: 14.14 -> 14.99 | Improved |
| PLOT geometry vs OVM3D-Det PCA on pseudo-labels | PLOT · nuScenes-metric labels | AOE (all, near): 1.360 -> 0.536 | Improved |
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.
- MonoSOWA — MonoSOWA: Scalable Monocular 3D Object Detector Without Human Annotations (ICCV 2025) · better on 6 of 6 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
- Leveraging Temporal Cues for Semi-Supervised Multi-View 3D Object Detection (CVPR 2025 · similarity 0.89)
- TrackCraft3R: Repurposing Video Diffusion Transformers for Dense 3D Tracking (NeurIPS 2026 · similarity 0.88)
- UNION: Unsupervised 3D Object Detection using Object Appearance-based Pseudo-Classes (NeurIPS 2024 · similarity 0.88)
- M3SOT: Multi-Frame, Multi-Field, Multi-Space 3D Single Object Tracking (AAAI 2024 · similarity 0.88)
- SeMoLi: What Moves Together Belongs Together (CVPR 2024 · similarity 0.88)
- MonoDiff: Monocular 3D Object Detection and Pose Estimation with Diffusion Models (CVPR 2024 · similarity 0.88)
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.
