Eaglenet Large Specialist R1

--- license: other license_name: proprietary license_link: https://github.com/Sph3inz/EAGLE-Net tags: - object-detection - ultralytics - yolo11 - drone-detection - counter-uav library_name: ultralytics pipeline_tag: object-detection --- # EAGLE-Net large-drone specialist (`large-specialist-r1`) Vanilla **YOLO11s** detector for the EAGLE-Net **hybrid** pipeline. Complements the NWD+P2 small-object specialist (P1) for drones **larger than ~50 px**. Trained on filtered **SynDroneVision** large-drone boxes (see `scripts/filter_large_drones.py` and `eaglenet train large`). ## Usage ```python from ultralytics import YOLO model = YOLO("hf://Sph3inxz/eaglenet-large-specialist-r1/best.pt") results = model.predict("video.mp4", imgsz=640, conf=0.30) ``` **Hybrid inference** (with P1 small specialist): ```bash eaglenet infer demo \ --weights <p1-best.pt> \ --weights-large hf://Sph3inxz/eaglenet-large-specialist-r1/best.pt \ --imgsz-large 640 --conf-large 0.30 \ --source clip.mp4 --out tracked.mp4 ``` ## Files | File | Description | |------|-------------| | `best.pt` | Ultralytics checkpoint (YOLO11s, 1-class drone) | ## Training (summary) - Data: SynDroneVision (large boxes only, filtered from full corpus) - Base: `yolo11s.pt` (COCO), stock CIoU, no P2 head - `imgsz=640`, scale aug `0.5`, single class `drone` - Intended for **large** boxes; poor Anti-UAV v4 scores are expected (different regime from P1) ## Source [EAGLE-Net](https://github.com/Sph3inz/EAGLE-Net)