YOLO Face Person Detector
iitolstykh/YOLO-Face-Person-Detector
published Nov 2025 · updated Nov 2025
YOLO Face Person Detector is a detection model fine-tuned from YOLOv8x to detect faces and persons in images.
specs
| Task | Object Detection |
| Architecture | YOLOv8x |
| Parameters | 68.2M (YOLOv8x base) |
| License | AGPL-3.0 |
about this model
iitolstykh/YOLO-Face-Person-Detector is a detection model that identifies faces and persons in images using a fine-tuned YOLOv8x architecture.
The model was trained on a proprietary dataset of approximately 150,000 images, enabling robust detection of two classes: Face and Person. The underlying YOLOv8x architecture provides high-capacity feature extraction, with the base model achieving a COCO mAP@50-95 of 53.9 and A100 TensorRT inference speed of 3.53 ms at 640px input size.
This fine-tuned model inherits the efficiency characteristics of the YOLOv8 family. Related work by the same authors (CerberusDet) demonstrates a 36% reduction in inference time compared to running individual models sequentially, suggesting efficient multi-class detection capability.
The model is licensed under AGPL-3.0. It is hosted as a managed API on Gigarouter, requiring no local installation — simply call the OpenAI-compatible endpoint for inference.
best for
- ·Detecting faces and persons in surveillance or security camera feeds
- ·Counting people and detecting faces in crowded event photos
FAQ
It detects two classes: Face and Person.
It is licensed under AGPL-3.0, inherited from Ultralytics YOLOv8.
Use the gigarouter OpenAI-compatible endpoint with your API key, specifying the model ID iitolstykh/YOLO-Face-Person-Detector.
It accepts images as URL, PIL.Image, or numpy array, with optional confidence and IoU thresholds.
It was fine-tuned from YOLOv8x on a proprietary dataset of approximately 150,000 images.
We're benchmarking and onboarding YOLO Face Person Detector as a hosted, OpenAI-compatible API. Sign in for free credit and be ready when it lands, or tell us you want it and we'll prioritize it.