Anime Eye Detector (YOLOv8)
A YOLOv8 object detection model fine-tuned to detect eyes in anime-style character art, intended for use with ComfyUI + Impact Pack for automated eye detailing/inpainting workflows (similar to how face_yolov8n.pt and hand_yolov8n.pt are used).
Model details (YOLOv8n and YOLOv8s)
- Base model:
yolov8n.pt/yolov8s.pt(Ultralytics) - Task: Object detection, single class (
eye) - Training data: A growing set of self-generated anime and semi-realistic style images (AI-generated, primarily female characters), manually labeled with bounding boxes around each visible eye. Currently 286 images (v2.0)
- Training config: 100 epochs, image size 640, batch size 8
Performance (on validation split)
| Metric | Nano (yolov8n) | Small (yolov8s) |
|---|---|---|
| mAP50 | 0.982 | 0.989 |
| mAP50-95 | 0.663 | 0.670 |
| Precision | 0.997 | 0.981 |
| Recall | 0.964 | 0.954 |
Known limitations
- Detection on male character eyes is improved as of v2.0 but still less reliable than on female characters.
- Trained on anime and semi-realistic style images — also tested well on realistic styles, though not extensively validated across every art style.
- Dataset size (286 images as of v2.0) — strong validation metrics, but a larger dataset would improve robustness further.
If you find specific failure cases, feel free to open a discussion — this is a good candidate for community-driven dataset expansion over time.
Examples
Detection preview — the model correctly finds eyes across different poses/styles:
Eye color change/Eye fixing — using the detected eye region with Detailer (SEGS) to redraw eye color/detail from a prompt, while keeping the rest of the image untouched:
Usage (ComfyUI)
Easiest method: Search "eye" in ComfyUI-Manager's Install Models menu and install directly — no manual download needed.
Manual method:
- Download the size you want (
eye_yolov8n.ptoreye_yolov8s.pt) from the Files tab. - Rename it if you want and place it in:
ComfyUI/models/ultralytics/bbox/ - Restart ComfyUI.
- In your workflow:
Load Image → UltralyticsDetectorProvider (select the eye detector model) → BboxDetectorSEGS → Detailer (SEGS) - Recommended
Detailer (SEGS)starting settings for eye detailing:guide_size: 512denoise: 0.5–0.7 (lower = closer to the original eye, higher = more prompt-driven reinterpretation)feather: 5–10
License
Released under the MIT License. Training images were self-generated by the author; users should independently verify licensing terms of any base checkpoint used to generate their own training/inference images if that matters for their use case.