Research map / Detection and segmentation / CNN features
Constellation Dataset: Benchmarking High-Altitude Object Detection for an Urban Intersection
Mehmet Kerem Türkcan, Chengbo Zang, Sanjeev Narasimhan, Gyung Hyun Je, Bo Yu, Mahshid Ghasemi, Javad Ghaderi, Gil Zussman, Zoran Kostić
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 |
|---|---|---|---|
| train at 2x input resolution (bilinear upscaling) | YOLOv8x · Constellation test | mAP@0.5: 93.0 -> 94.8 | Improved |
| add Real-ESRGAN super-resolution to 2x training | YOLOv8x (2x) · Constellation test | mAP@0.5: 94.8 -> 95.0 | Improved |
| add Real-ESRGAN super-resolution to 2x training | YOLOv8n (2x) · Constellation test | mAP@0.5: 94.8 -> 94.2 | Hurt |
| use P2-P6 detection layers instead of default P3-P5 | YOLOv8x · Constellation test | mAP@0.5: 93.0 -> 94.3 | Improved |
| add P2 detection layer | YOLOv8x · Constellation test | mAP@0.5: 93.0 -> 94.0 | Improved |
| add P6 detection layer | YOLOv8x · Constellation test | mAP@0.5: 93.0 -> 94.0 | Improved |
| pretrain on VisDrone in addition to COCO (ablation) | YOLOv8x · Constellation test (no augmentation) | mAP@0.5: 93.0 -> 94.8 | Improved |
| pretrain on CARLA synthetic data in addition to COCO (ablation) | YOLOv8x · Constellation test (with augmentation) | mAP@0.5: 94.7 -> 94.4 | Hurt |
| pretrain on Stanford Drone Dataset in addition to COCO (ablation) | YOLOv8x · Constellation test (no augmentation) | mAP@0.5: 93.0 -> 93.0 | No clear effect |
| use pretraining (COCO) instead of random initialization with augmentation (ablation) | YOLOv8x · Constellation test (with augmentation) | mAP@0.5: 94.7 vs 94.7 | No clear effect |
| add custom weather augmentations (MotionBlur, RandomShadow, RandomRain) (ablation) | YOLOv8x COCO · Constellation test | mAP@0.5: 93.0 -> 94.7 | Improved |
| add custom augmentations to VisDrone-pretrained model (ablation) | YOLOv8x COCO+VisDrone · Constellation test | mAP@0.5: 94.8 -> 95.4 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- YOLOv7 — YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors (CVPR 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Object Detection in Aerial Images: A Large-Scale Benchmark and Challenges (TPAMI 2021 · similarity 0.87)
- Towards Large-Scale Small Object Detection: Survey and Benchmarks (TPAMI 2023 · similarity 0.87)
- Benchmarking Object Detectors under Real-World Distribution Shifts in Satellite Imagery (CVPR 2025 · similarity 0.86)
- Generalizable Pedestrian Detection: The Elephant in the Room (CVPR 2021 · similarity 0.86)
- COUNTS: Benchmarking Object Detectors and Multimodal Large Language Models under Distribution Shifts (CVPR 2025 · similarity 0.84)
- Roboflow100-VL: A Multi-Domain Object Detection Benchmark for Vision-Language Models (NeurIPS 2025 · similarity 0.84)
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.
