Research map / Detection and segmentation / CNN features
TIDE: A General Toolbox for Identifying Object Detection Errors
Daniel Bolya, Sean Foley, James Hays, Judy Hoffman
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 mask scoring (ablation) | Mask R-CNN · COCO box AP50 | AP50: 58.1 -> 58.3 | Improved |
| add mask scoring (ablation) | Mask R-CNN · COCO | localization error dAP loc: 9.3 -> 7.8 | Improved |
| add mask scoring (ablation) | Mask R-CNN · COCO | classification error dAP cls: 3.1 -> 3.6 | Hurt |
| add mask scoring (ablation) | Mask R-CNN · COCO | background error dAP bkg: 4.5 -> 5.1 | Hurt |
| add mask scoring (ablation) | YOLACT++ · COCO | AP50: 51.8 -> 52.3 | Improved |
| add mask scoring (ablation) | YOLACT++ · COCO | localization error dAP loc: 10.4 -> 9.7 | Improved |
| add mask scoring (ablation) | YOLACT++ · COCO | false negative error dAP FN: 27.7 -> 28.2 | Hurt |
| add mask scoring (ablation) | YOLACT++ · COCO | false positive error dAP FP: 10.7 -> 10.1 | Improved |
| raise foreground IoU threshold from 0.5 to 0.9 (ablation) | Mask R-CNN · COCO | localization error dAP loc: 6.2 -> 55.1 | Hurt |
| raise foreground IoU threshold from 0.5 to 0.9 (ablation) | Mask R-CNN · COCO | AP: 61.7 -> 12.0 | Hurt |
| raise foreground IoU threshold from 0.5 to 0.9 (ablation) | Mask R-CNN · COCO | false negative error dAP FN: 15.3 -> 33.2 | Hurt |
| compute errors progressively instead of individually (swap classification and background oracle order) (ablation) | Mask R-CNN · COCO AP50 | perceived background error | Hurt |
Closest papers by meaning
- Object Detection Benchmarks are Incomplete: The Role of Label Errors and Annotation Uncertainty (NeurIPS 2026 · similarity 0.86)
- ObjectNet: A large-scale bias-controlled dataset for pushing the limits of object recognition models (NeurIPS 2019 · similarity 0.85)
- What Ails One-Shot Image Segmentation: A Data Perspective (NeurIPS 2021 · similarity 0.85)
- A Simple Framework for Open-Vocabulary Segmentation and Detection (ICCV 2023 · similarity 0.85)
- The Devil is in the Object Boundary: Towards Annotation-free Instance Segmentation using Foundation Models (ICLR 2024 · similarity 0.84)
- How to Evaluate the Generalization of Detection? A Benchmark for Comprehensive Open-Vocabulary Detection (AAAI 2024 · similarity 0.83)
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.
