Research map / Detection and segmentation / CNN features
Hit-Detector: Hierarchical Trinity Architecture Search for Object Detection
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Search backbone, neck, and head together end-to-end | Hit-Detector · COCO minival | mAP: 36.2 -> 41.4 | Helped |
| Combine separately searched DetNAS backbone and NAS-FPN neck | DetNAS-1.3G + NASFPN · COCO minival | mAP: 40.2 -> 39.4 | Hurt |
| Use separate component-specific sub search spaces instead of one shared space (ablation) | Hit-Detector · COCO minival | mAP: 40.1 -> 41.4 | Helped |
| Replace the baseline backbone with the searched backbone (ablation) | FPN baseline · COCO minival | mAP: 36.2 -> 39.2 | Helped |
| Replace the baseline neck with the searched neck (ablation) | FPN baseline · COCO minival | mAP: 36.2 -> 37.4 | Helped |
| Replace the baseline head with the searched head (ablation) | FPN baseline · COCO minival | mAP: 36.2 -> 38.5 | Helped |
| Search all three detector components rather than only one component (ablation) | Hit-Detector · COCO minival | mAP: 39.2 -> 41.4; 37.4 -> 41.4; 38.5 -> 41.4 | Helped |
| Add column-sparse regularization during sub-space screening (ablation) | Hit-Detector screening · Operation screening | operation-score separation | Helped |
| Use no column-sparse regularization (μ=0) (ablation) | Hit-Detector screening · Operation screening | screening stability | Hurt |
| Apply Hit-Detector search to RetinaNet | RetinaNet · RetinaNet COCO | mAP: 35.6 -> 36.9 | Helped |
Closest papers by meaning
- SM-NAS: Structural-to-Modular Neural Architecture Search for Object Detection (AAAI 2020 · similarity 0.92)
- SP-NAS: Serial-to-Parallel Backbone Search for Object Detection (CVPR 2020 · similarity 0.90)
- NAS-FCOS: Fast Neural Architecture Search for Object Detection (CVPR 2020 · similarity 0.89)
- Joint-DetNAS: Upgrade Your Detector With NAS, Pruning and Dynamic Distillation (CVPR 2021 · similarity 0.88)
- DetNAS: Backbone Search for Object Detection (NeurIPS 2019 · similarity 0.88)
- Can Weight Sharing Outperform Random Architecture Search? An Investigation With TuNAS (CVPR 2020 · similarity 0.87)
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.