Anzhcs_YOLOs
Anzhc/Anzhcs_YOLOs
A popular open object detection model, with 75.5K downloads a month. gigarouter benchmarks and hosts it as an OpenAI-compatible API.
about this model
This collection of YOLO models provides specialized object detection and segmentation for faces, eyes, head+hair, and breasts. All models are trained on custom-annotated datasets and are hosted by gigarouter as an OpenAI-compatible API.
Face Segmentation
Universal models (v2–v4) work on both illustrations and real faces. The latest v4 y11n model achieves mAP50 0.835 (box) and 0.800 (mask) at 640px. Gendered real-face models are available: ManFace (mAP50 0.883 box/mask) and WomanFace (mAP50 0.82 box/mask).
Eyes Segmentation
Designed for inpainting anime eyes, detecting sclera area. Achieves mAP50 0.925 (box) and 0.868 (mask) at 1024px.
Head+Hair Segmentation
Detects head and hair for illustration and real images. Y8n: mAP50 0.775 (box), 0.777 (mask); Y8m: 0.867 (box), 0.862 (mask) at 640px.
Breast Segmentation and Size Detection
Segmentation models are trained on anime images (e.g., v1 1024m: mAP50 0.782 box, 0.775 mask). A breast size detection/classification model (v8, 15 classes) provides accuracy metrics including +-1 class accuracy, miss ratio, and false-positive ratio.
We're benchmarking and onboarding Anzhcs_YOLOs 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.