Research map / Detection and segmentation / CNN features
YOLOV: Making Still Image Object Detectors Great at Video Object Detection
Read the paperCheck 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 |
|---|---|---|---|
| Add temporal feature aggregation with affinity-guided confidence weighting (ablation) | YOLOX-S on ImageNet VID · ImageNet VID val | AP50 (%): 69.5 -> 76.9 | Helped |
| Add average pooling over reference features to affinity-based aggregation (ablation) | YOLOV-S on ImageNet VID · ImageNet VID val | AP50 (%): 76.9 -> 77.3 | Helped |
| Use global rather than local reference-frame sampling (ablation) | YOLOV · ImageNet VID val, 39 reference frames | AP50 (%): 73.6 -> 77.3 | Helped |
| Increase global reference frames from 3 to 31 (ablation) | YOLOV · ImageNet VID val | AP50 (%): 74.6 -> 77.3 | Helped |
| Increase local reference frames from 3 to 39 (ablation) | YOLOV · ImageNet VID val | AP50 (%): 70.4 -> 73.6 | Helped |
| Increase FSM proposals per frame from 10 to 75 (ablation) | YOLOV · ImageNet VID val | AP50 (%): 76.1 -> 77.4 | Helped |
| Increase average-pooling threshold from τ=0 to τ=0.75 (ablation) | YOLOV · ImageNet VID val | AP50 (%): 73.1 -> 77.3 | Helped |
| Raise average-pooling threshold to τ=1 (ablation) | YOLOV · ImageNet VID val | AP50 (%): 77.3 -> 76.9 | Hurt |
| Add strong augmentation during fine-tuning (ablation) | YOLOV-X · ImageNet VID val | AP50 (%): 85.0 -> 85.5 | Helped |
| Apply the YOLOV strategy to PPYOLOE-S | PPYOLOE-S · ImageNet VID val | AP50 (%): 69.5 -> 74.9 | Helped |
| Apply the YOLOV strategy to PPYOLOE-L | PPYOLOE-L · ImageNet VID val | AP50 (%): 76.9 -> 82.0 | Helped |
| Apply the YOLOV strategy to FCOS | FCOS · ImageNet VID val | AP50 (%): 67.0 -> 73.1 | Helped |
Closest papers by meaning
- YOLOv10: Real-Time End-to-End Object Detection (NeurIPS 2024 · similarity 0.89)
- YOLO9000: Better, Faster, Stronger (CVPR 2017 · similarity 0.88)
- You Only Look Once: Unified, Real-Time Object Detection (CVPR 2016 · similarity 0.88)
- YOLOv12: Attention-Centric Real-Time Object Detectors (NeurIPS 2025 · similarity 0.85)
- YOLO-RD: Introducing Relevant and Compact Explicit Knowledge to YOLO by Retriever-Dictionary (ICLR 2025 · similarity 0.85)
- SimLTD: Simple Supervised and Semi-Supervised Long-Tailed Object Detection (CVPR 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.