Research map / Detection and segmentation / Weakly supervised detection and segmentation
Semi-Supervised 3D Object Detection with Proficient Teachers
Junbo Yin, Jin Fang, Dingfu Zhou, Liangjun Zhang, Cheng-Zhong Xu, Jianbing Shen, Wenguan Wang
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 spatial-temporal ensemble (STE) to the teacher (ablation) | Noisy Student (c=0.3), SECOND · ONCE val (Small) | mAP: 55.75 -> 56.21 (+0.46) | Improved |
| add box-wise contrastive learning (BCL) (ablation) | Noisy Student (c=0.3), SECOND · ONCE val (Small) | mAP: 55.75 -> 56.54 (+0.79) | Improved |
| replace confidence thresholding with clustering-based box voting (CBV) on top of STE (ablation) | STE teacher, SECOND · ONCE val (Small) | mAP: 55.75 -> 57.17 (+1.42) | Improved |
| add BCL to STE+CBV (full model) (ablation) | ProficientTeachers, SECOND · ONCE val (Small) | mAP: 57.17 -> 57.72 | Improved |
| add STE to the vanilla teacher without filtering (ablation) | SECOND teacher · Vehicle pseudo labels | recall / precision: recall 78.3% -> 83.9%, precision 47.2% -> 27.4% | Mixed |
| add CBV after STE (ablation) | SECOND + STE · Vehicle pseudo labels | precision: 27.4% -> 76.5%; recall 83.9% -> 83.6% | Improved |
| replace CBV with confidence threshold c=0.3 after STE (ablation) | SECOND + STE · Vehicle pseudo labels | recall / precision: CBV 83.6% / 76.5% vs threshold 80.4% / 74.6% | Hurt |
| add ProficientTeachers SSL framework | SECOND · ONCE val (Small/Medium/Large) | mAP: +5.83, +8.00, +9.51 mAP | Improved |
| add ProficientTeachers SSL framework | PV-RCNN · ONCE val (Medium) | mAP: 57.24 -> 63.40 (+6.16) | Improved |
| add ProficientTeachers SSL framework | CenterPoint · ONCE val (Medium) | mAP: 62.99 -> 68.22 (+5.23) | Improved |
| use Pseudo Label SSL without student augmentation | SECOND · ONCE val (Small/Medium/Large) | mAP: 51.89 -> 51.22, 50.40, 49.76 | Hurt |
| increase amount of unlabeled data from Small to Large | ProficientTeachers, SECOND · ONCE val | mAP: 57.72 -> 59.89 -> 61.40 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- 3DIoUMatch — 3DIoUMatch: Leveraging IoU Prediction for Semi-Supervised 3D Object Detection (CVPR 2021) · better on 4 of 4 shared results
- SESS — SESS: Self-Ensembling Semi-Supervised 3D Object Detection (CVPR 2020) · better on 4 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
- UpCycling: Semi-supervised 3D Object Detection without Sharing Raw-level Unlabeled Scenes (ICCV 2023 · similarity 0.91)
- Leveraging Temporal Cues for Semi-Supervised Multi-View 3D Object Detection (CVPR 2025 · similarity 0.90)
- Semi-supervised 3D Object Detection with PatchTeacher and PillarMix (AAAI 2024 · similarity 0.89)
- A-Teacher: Asymmetric Network for 3D Semi-Supervised Object Detection (CVPR 2024 · similarity 0.88)
- Leveraging Imagery Data with Spatial Point Prior for Weakly Semi-supervised 3D Object Detection (AAAI 2024 · similarity 0.88)
- Dual-Perspective Knowledge Enrichment for Semi-supervised 3D Object Detection (AAAI 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.
