Research map / Point clouds / Segmentation and detection
MixCycle: Mixup Assisted Semi-Supervised 3D Single Object Tracking with Cycle Consistency
Qiao Wu, Jiaqi Yang, Kun Sun, Chu'ai Zhang, Yanning Zhang, Mathieu Salzmann
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 MixCycle semi-supervised training with unlabeled data to P2B at 1% labels | P2B · KITTI (1% labels) | Precision: 5.0 -> 33.4 | Improved |
| add MixCycle to P2B using 10% labels versus fully supervised P2B with 100% labels | P2B · KITTI | Success: 42.4 -> 43.8 | Improved |
| MixCycle with 10% labels versus fully supervised BAT with 100% labels | BAT · KITTI | Success: 51.2 -> 46.2 | Hurt |
| add MixCycle with all labels | BAT · KITTI Car, 100% labels | Success: 60.5 -> 64.7 | Improved |
| add SOTMixup to the cycle tracking framework (ablation) | MixCycle · KITTI 10% labels | Success: 38.8 -> 46.2 | Improved |
| use forward-backward cycle instead of only self cycle (ablation) | MixCycle · KITTI 10% labels | Success: 34.9 -> 38.4 | Improved |
| combine self cycle with forward-backward cycle (ablation) | MixCycle · KITTI 10% labels | Success: 38.4 -> 39.5 | Improved |
| add random rigid transformation without SOTMixup (ablation) | MixCycle · KITTI 10% labels | Success: 39.5 -> 38.8 | Hurt |
| add random rigid transformation together with SOTMixup (ablation) | MixCycle · KITTI 10% labels | Success: 44.6 -> 46.2 | Improved |
| use 3-step forward-backward cycle instead of 2 steps (ablation) | MixCycle · KITTI 10% labels | Success: 45.8 -> 46.2 | Improved |
| apply SOTMixup to the search area only instead of also template (ablation) | MixCycle · KITTI 10% labels | Success: 44.5 -> 46.2 | Improved |
| apply SOTMixup also in backward tracking (ablation) | MixCycle · KITTI 10% labels | Success: 46.2 -> 44.7 | Hurt |
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.
- MLVSNet — MLVSNet: Multi-Level Voting Siamese Network for 3D Visual Tracking (ICCV 2021) · better on 2 of 2 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
- M3SOT: Multi-Frame, Multi-Field, Multi-Space 3D Single Object Tracking (AAAI 2024 · similarity 0.88)
- S2-Track: A Simple yet Strong Approach for End-to-End 3D Multi-Object Tracking (ICML 2025 · similarity 0.88)
- Towards Category Unification of 3D Single Object Tracking on Point Clouds (ICLR 2024 · similarity 0.87)
- Modeling Continuous Motion for 3D Point Cloud Object Tracking (AAAI 2024 · similarity 0.86)
- Leveraging Temporal Cues for Semi-Supervised Multi-View 3D Object Detection (CVPR 2025 · similarity 0.86)
- MixSup: Mixed-grained Supervision for Label-efficient LiDAR-based 3D Object Detection (ICLR 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.
