dronefreak/seadronessee-rfdetr-medium

RF-DETR Medium Finetuned on SeaDronesSee

Fine-tuned RF-DETR Medium object detector on the SeaDronesSee 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.

---

Detection Showcase

---

Performance

Metric	Score (%)
mAP@50	83.47
mAP@50-95	47.49
Precision	87.01
Recall	83.33
F1 Score	85.13
Parameters	33.7M
FLOPs	N/A (not published upstream)

---

Evaluation Protocol

Metrics reported in this model card are computed on the SeaDronesSee val split, using DetectionBench's standard evaluation pipeline (detectionbench-evaluate).

---

SeaDronesSee Model Zoo

Every model DetectionBench has trained and evaluated on SeaDronesSee 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	83.47	47.49	87.01	83.33
YOLOv26m	82.38	49.57	90.01	81.18
RF-DETR Small	80.97	45.31	85.68	80.16
YOLOv26s	80.14	47.35	88.5	77.51
YOLOv11x	74.82	45.56	87.37	72.46
YOLOv8s	72.94	43.05	84.52	71.25
RF-DETR Nano	72.38	39.83	81.37	74.08
YOLOv11n	69.93	40.41	82.87	69.04
YOLOv8n	69.22	40.35	82.46	68.36
YOLOv8m	62.08	34.41	77.3	61.01

---

Per-Class Performance

Class	mAP@50	mAP@50-95
swimmer	76.06	30.44
boat	95.71	69.35
jetski	93.55	64.16
life_saving_appliances	70.55	25.24
buoy	81.49	48.26

---

Evaluation Visualizations

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 SeaDronesSee. For the full dataset description, provenance, license, and citation, see the dataset card:

https://huggingface.co/datasets/dronefreak/SeaDronesSee

Classes

• swimmer
• boat
• jetski
• life_saving_appliances
• buoy

---

Usage

Install Dependencies

pip install rfdetr huggingface_hub

Load Model from Hugging Face

from huggingface_hub import hf_hub_download
import rfdetr

weights = hf_hub_download(
    repo_id="dronefreak/seadronessee-rfdetr-medium",
    filename="checkpoint_best_total.pth"
)

model = rfdetr.RFDETRMedium(pretrain_weights=weights)

Run Inference

detections = model.predict("image.jpg", threshold=0.25)

---

Training Configuration

Setting	Value
Dataset	SeaDronesSee
Framework	RF-DETR
Training Toolkit	DetectionBench
Epochs (configured max)	500
Epochs (actually trained)	123
Early Stopping Patience	100
Batch Size	5
Resolution	576
Optimizer	adamw
Learning Rate	0.0001
Seed	42

---

Repository Contents

checkpoint_best_total.pth
metrics.csv
config.json
seadronessee_rfdetr-medium_showcase.jpg
README.md

---

Related Resources

• SeaDronesSee dataset card on Hugging Face
• 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)
• Hardware profiling (latency, FPS, VRAM, parameters, FLOPs)
• One-command reproducibility via versioned Hydra configs

If you find this model useful, please consider starring the repository.

---

Known Limitations

• Severe class imbalance: swimmer (64.22%) and boat (22.55%) account for roughly 87% of all annotated boxes in the training set, while life_saving_appliances (1.60%) is rare -- per-class accuracy on the minority classes is measured on comparatively few examples.
• Small-object heavy: objects are captured from altitude over open water, so roughly 79% of boxes cover under 0.1% of the image area -- swimmers and buoys in particular are small, low-contrast targets against water.
• No public test-split labels: the official images/test/ split is a held-out competition set with no released ground truth, so these models are evaluated on the valid split instead of test -- the number reported here is not directly comparable to official SeaDronesSee leaderboard submissions, which score against the held-out test set via the benchmark's own server.
• A maritime search-and-rescue benchmark specifically: generalization to non-maritime aerial scenes, different water/lighting conditions, or altitudes outside this dataset's capture range is untested.

---

Citation

If you use this model in your research, please consider citing:

1. The SeaDronesSee dataset (see below)
2. The original RF-DETR Medium architecture (see below)
3. The other model architectures shown in the Model Zoo/External Comparison tables above, if you reference their results
4. DetectionBench, the training/evaluation framework used to produce this checkpoint

@inproceedings{varga2022seadronessee,
  title={SeaDronesSee: A maritime benchmark for detecting humans in open water},
  author={Varga, Leon Amadeus and Kiefer, Benjamin and Messmer, Martin and Zell, Andreas},
  booktitle={Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},
  pages={2260--2270},
  year={2022}
}

@misc{varga2021seadronesseemaritimebenchmarkdetecting,
      title={SeaDronesSee: A Maritime Benchmark for Detecting Humans in Open Water},
      author={Leon Amadeus Varga and Benjamin Kiefer and Martin Messmer and Andreas Zell},
      year={2021},
      eprint={2105.01922},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2105.01922}
}

@inproceedings{robinson2026rfdetr,
  title     = {RF-DETR: Real-Time Detection Transformer},
  author    = {Robinson, Isaac and Robicheaux, Peter and Popov, Matvei and Ramanan, Deva and Peri, Neehar},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year      = {2026},
  url       = {https://arxiv.org/abs/2511.09554}
}

@article{oquab2023dinov2,
  title={DINOv2: Learning Robust Visual Features without Supervision},
  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},
  journal={arXiv preprint arXiv:2304.07193},
  year={2023}
}

Other architectures compared against on SeaDronesSee in this model card:

YOLOv11

No official YOLO11 research paper has been published by Ultralytics; the most commonly cited independent architectural analysis is used instead:

@article{khanam2024yolov11,
  title={YOLOv11: An Overview of the Key Architectural Enhancements},
  author={Khanam, Rahima and Hussain, Muhammad},
  journal={arXiv preprint arXiv:2410.17725},
  year={2024}
}

YOLOv26

@article{jocher2026yolo26,
  title={Ultralytics YOLO26: Unified Real-Time End-to-End Vision Models},
  author={Jocher, Glenn and Qiu, Jing and Liu, Mengyu and Lyu, Shuai and Akyon, Fatih Cagatay and Kalfaoglu, Muhammet Esat},
  journal={arXiv preprint arXiv:2606.03748},
  year={2026}
}

YOLOv8

No official YOLOv8 research paper has been published by Ultralytics; this is their own recommended software citation instead:

@software{jocher2023yolov8,
  author = {Glenn Jocher and Ayush Chaurasia and Jing Qiu},
  title = {Ultralytics YOLOv8},
  version = {8.0.0},
  year = {2023},
  url = {https://github.com/ultralytics/ultralytics},
  license = {AGPL-3.0}
}

@software{Saksena_DetectionBench_2026,
  author = {Saksena, Saumya Kumaar},
  title = {DetectionBench: Reproducible Benchmarks for Modern Object Detectors on Real-World Datasets},
  url = {https://github.com/dronefreak/DetectionBench},
  year = {2026}
}