Fine-tuned RF-DETR Small object detector on the VisDrone-DET benchmark dataset, trained and evaluated as part of DetectionBench -- a framework for reproducibly benchmarking modern object detectors with identical training recipes and evaluation metrics across multiple real-world datasets.
Metrics reported in this model card are computed on the VisDrone-DET test split, using DetectionBench's standard evaluation pipeline (detectionbench-evaluate).
VisDrone-DET Model Zoo
Every model DetectionBench has trained and evaluated on VisDrone-DET so far, for full transparency -- see DetectionBench for the smaller, curated comparison set used on the project README.
Model
mAP@50
mAP@50-95
Precision
Recall
RF-DETR Medium
36.82
20.14
64.0
47.05
RF-DETR Small
33.25
17.88
62.62
43.51
RF-DETR Nano
25.15
12.77
58.99
35.0
External VisDrone-DET Comparison
The YOLO/RT-DETR rows below were trained and evaluated on VisDrone2019-DET's test split via a separate companion codebase (VisDrone-dataset-python-toolkit), not reproduced inside DetectionBench -- included here purely for context. The RF-DETR rows are this repository's own DetectionBench-trained runs (see the Model Zoo table above).
This model was evaluated with Supervision's detection metrics, which report mAP/Precision/Recall directly but don't produce PR-curve, F1-curve, or confusion-matrix plot images the way Ultralytics' validator does. See the Performance table above for Precision/Recall/F1 and the per-class table above for the full per-class mAP breakdown.
Dataset
This model was trained on VisDrone-DET. For the full dataset description, provenance, license, and citation, see the dataset card:
DetectionBench -- reproducible benchmarks for modern object detectors on real-world datasets
Training Framework
This model was trained using DetectionBench, an open-source framework for benchmarking object detectors across multiple real-world datasets with a common pipeline.
Features include:
A dataset-adapter registry for converting real-world datasets into a canonical format
Identical training/evaluation recipes across model families (Ultralytics YOLO/RT-DETR, RF-DETR)
One-command reproducibility via versioned Hydra configs
If you find this model useful, please consider starring the repository.
Known Limitations
Severe class imbalance: car (42.21%) and pedestrian (23.12%) account for two-thirds of all annotated boxes in the training set, while awning-tricycle (0.95%) and tricycle (1.40%) are rare -- the others class has zero annotated instances in the training set entirely and is effectively unusable (always 0 AP).
Extreme small-object density: ~53 annotated boxes per image on average, with roughly 69% of boxes covering under 0.1% of the image area -- consistent with VisDrone's aerial small-object detection challenge (objects captured from significant altitude).
The original authors license VisDrone under CC BY-NC-SA 3.0 -- non-commercial research use only (see the dataset's homepage); this applies to any model trained on it, not only the raw images.
These RF-DETR checkpoints were trained/evaluated directly through DetectionBench. The YOLO/RT-DETR rows in the External VisDrone Model Zoo comparison below were trained via a separate companion codebase, not reproduced inside DetectionBench -- see that collection for their own training details and caveats.
Citation
If you use this model in your research, please consider citing:
The VisDrone-DET dataset (see below)
The original RF-DETR Small architecture (see below)
DetectionBench, the training/evaluation framework used to produce this checkpoint
@article{zhu2018vision,
title={Vision meets drones: A challenge},
author={Zhu, Pengfei and Wen, Longyin and Bian, Xiao and Ling, Haibin and Hu, Qinghua},
journal={arXiv preprint arXiv:1804.07437},
year={2018}
}
bibtex
1@inproceedings{robinson2026rfdetr,
2 title = {RF-DETR: Real-Time Detection Transformer},
3 author = {Robinson, Isaac and Robicheaux, Peter and Popov, Matvei and Ramanan, Deva and Peri, Neehar},
4 booktitle = {International Conference on Learning Representations (ICLR)},
5 year = {2026},
6 url = {https://arxiv.org/abs/2511.09554}
7}
89@article{oquab2023dinov2,
10 title={DINOv2: Learning Robust Visual Features without Supervision},
11 author={Oquab, Maxime and Darcet, Timoth{\'e}e and Moutakanni, Theo and Vo, Huy and Szafraniec, Marc and Khalidov, Vasil and Fernandez, Pierre and Haziza, Daniel and Massa, Francisco and El-Nouby, Alaaeldin and others},
12 journal={arXiv preprint arXiv:2304.07193},
13 year={2023}
14}
bibtex
1@software{Saksena_DetectionBench_2026,
2 author = {Saksena, Saumya Kumaar},
3 title = {DetectionBench: Reproducible Benchmarks for Modern Object Detectors on Real-World Datasets},
4 url = {https://github.com/dronefreak/DetectionBench},
5 year = {2026}
6}
