Research map / Detection and segmentation / CNN features
You Only Look One-Level Feature
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 |
|---|---|---|---|
| Use a single-level C5 detector with Dilated Encoder and Uniform Matching instead of RetinaNet+'s feature pyramid | YOLOF vs RetinaNet+ · COCO2017 val | FPS: 13 -> 32 | Helped |
| Use YOLOF's single-level detector rather than DETR | YOLOF and DETR, ResNet-50 · COCO2017 val | AP: 42.0 -> 41.6 | Hurt |
| Use YOLOF's single-level detector rather than DETR | YOLOF-R101 and DETR-R101 · COCO2017 val | AP: 43.5 -> 43.7 | Helped |
| Use YOLOF-DC5 with the reported training and augmentation recipe rather than YOLOv4 | YOLOF-DC5 vs YOLOv4 · COCO test-dev | AP: 43.5 -> 44.3 | Helped |
| Add Dilated Encoder to the single-level detector (ablation) | YOLOF · COCO2017 val | AP: 21.1 -> 29.1 | Helped |
| Add Uniform Matching to the single-level detector (ablation) | YOLOF · COCO2017 val | AP: 21.1 -> 33.8 | Helped |
| Add both Dilated Encoder and Uniform Matching (ablation) | YOLOF · COCO2017 val | AP: 21.1 -> 37.7 | Helped |
| Increase the number of encoder residual blocks from 0 to 10 (ablation) | YOLOF · COCO2017 val | AP: 33.8 -> 36.9 | Helped |
| Use progressively varied dilation rates 2,4,6,8 rather than 1,1,1,1 (ablation) | YOLOF · COCO2017 val | AP: 35.5 -> 37.7 | Helped |
| Increase dilation rates further from 2,4,6,8 to 3,6,9,12 (ablation) | YOLOF · COCO2017 val | AP: 37.7 -> 37.3 | Hurt |
| Add residual shortcuts to the encoder blocks with dilation rates 2,4,6,8 (ablation) | YOLOF · COCO2017 val | AP: 34.1 -> 37.7 | Helped |
| Increase Uniform Matching from top1 to top4 anchors per ground-truth box (ablation) | YOLOF · COCO2017 val | AP: 35.9 -> 37.7 | Helped |
Closest papers by meaning
- M2Det: A Single-Shot Object Detector Based on Multi-Level Feature Pyramid Network (AAAI 2019 · similarity 0.90)
- Feature Pyramid Networks for Object Detection (CVPR 2017 · similarity 0.89)
- AugFPN: Improving Multi-Scale Feature Learning for Object Detection (CVPR 2020 · similarity 0.88)
- NAS-FPN: Learning Scalable Feature Pyramid Architecture for Object Detection (CVPR 2019 · similarity 0.87)
- Learning 1-Bit Tiny Object Detector with Discriminative Feature Refinement (ICML 2024 · similarity 0.86)
- HS-FPN: High Frequency and Spatial Perception FPN for Tiny Object Detection (AAAI 2025 · similarity 0.85)
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.