Research map / Detection and segmentation / Weakly supervised detection and segmentation
Mind the Gap: Transferring Labels to Align Object Detection Datasets
Mikhail Kennerley, Angelica I. Aviles-Rivero, Carola-Bibiane Schönlieb, Robby T. Tan
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 class-aware SFF attention instead of no attention (ablation) | LAT (FRCNN) · Cityscapes | AP: 57.5 -> 60.1 | Improved |
| use class-aware SFF instead of standard attention (ablation) | LAT (FRCNN) · Waymo | AP: 47.0 -> 48.5 | Improved |
| apply scaling to class branch rather than feature branch (ablation) | LAT (FRCNN) · Cityscapes | AP: 59.6 -> 60.1 | Improved |
| add attention scaling versus none (ablation) | LAT (FRCNN) · Cityscapes | AP: 57.8 -> 60.1 | Improved |
| fine-tune on source-only batches after mixed pretraining (ablation) | LAT downstream detector · Cityscapes | AP: 58.9 -> 60.1 | Improved |
| 50/50 batches instead of mixed batches (ablation) | LAT downstream detector · Cityscapes | AP: 58.9 -> 57.5 | Hurt |
| add SAM3 predictions to LAT | LAT (RT-DETR) · Cityscapes | AP: 60.6 -> 61.0 | Improved |
| use SAM3 labels alone instead of LAT | label transfer · nuImages | AP: 39.5 -> 32.6 | Hurt |
| semi-supervised student-teacher using other datasets | FRCNN baseline · Cityscapes | AP: 55.2 -> 55.1 | No clear effect |
| LAT on large target datasets with standard training | FRCNN · BDD100K | AP: 57.2 -> 56.1 | Hurt |
| longer training for LAT | FRCNN · BDD100K | AP: 56.1 -> 57.8 | Improved |
| longer training for LAT | RT-DETR · Cityscapes | AP: 61.2 -> 59.8 | Hurt |
Closest papers by meaning
- Transferring Labels to Solve Annotation Mismatches Across Object Detection Datasets (ICLR 2024 · similarity 0.94)
- Object Detection Benchmarks are Incomplete: The Role of Label Errors and Annotation Uncertainty (NeurIPS 2026 · similarity 0.90)
- ScaleDet: A Scalable Multi-Dataset Object Detector (CVPR 2023 · similarity 0.89)
- Object Detection with a Unified Label Space from Multiple Datasets (ECCV 2020 · similarity 0.89)
- Beyond Boundaries: Leveraging Vision Foundation Models for Source-Free Object Detection (AAAI 2026 · similarity 0.88)
- [CLS] is Not Enough: Multi-Label Recognition via Patch-Level Inference and Adaptive Aggregation (ICML 2026 · 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.
