Rabbit GI health prototype
Your bunny's gut health, reviewed with care.
A polished AI product concept for spotting early litter-box warning signs, translating tiny poop-pattern changes into clear next steps for rabbit owners.
GI watch score
72
Phone-photo ready
Shop by warning sign
Small changes can mean big urgency.
The interface organizes model flags into owner-friendly categories, so a stressful litter-box photo becomes easier to understand.
Pellet size drops
Misshapen pellets
Hair strings
Cecotrope clusters
This prototype is for product and ML workflow design only. It is not a diagnosis. If a rabbit stops eating, stops producing droppings, or seems painful or lethargic, contact a rabbit-savvy veterinarian urgently.
Photo triage workflow
Upload a litter-box photo and review model flags.
The scan demo simulates how the app would turn object detections into a readable triage summary for owners and vet handoff.
Image input
Litter box scan
small pellets
0.86
misshapen
0.79
hair chain
0.71
cecotrope cluster
0.68
Meet the system
A gentle interface over a serious ML workflow.
Label taxonomy
Pellet size drop, misshapen pellet, hair string, cecotrope cluster, normal pellet, litter edge, and hay occlusion classes.
PyTorch detector
YOLOv11-ready training flow with augmentation for low light, hay clutter, varied litter substrates, and phone camera angles.
Health metrics
Aggregates detections into size distribution, shape irregularity, hair-chain count, and cecotrope-confidence signals.
Deployment plan
FastAPI inference, containerized for repeatable demos.
The included backend scaffold shows how trained weights would be served behind a small image upload endpoint, ready for Docker deployment and later mobile upload flows.
Stack
- Training: PyTorch + YOLO-style object detection
- Serving: FastAPI image upload endpoint
- Deployment: Docker container with pinned requirements
- UX: accessible triage UI with vet escalation copy