Portfolio case study
Pores & Wrinkles Detection Service
Face texture analysis service that detects pores and wrinkles and returns labeled overlays and metrics.
Overview
Cosmetic face-texture pipeline using face landmarks, region masks, segmentation-based wrinkle and fine-line detection, skeletonized line traces, overlays, per-line CSV outputs, timing events, and visual quality gates. The public case study avoids diagnostic claims and focuses on the engineering path from image capture to reviewable overlays.
What It Covers
- MediaPipe landmark-based ROI extraction and face-region masks
- Segmentation-based wrinkle and fine-line tracing with skeleton overlays
- Async job API with progress + results endpoints
- Flutter demo client and Telegram Mini App UI for cosmetic analysis review
Stack And Topics
- Python
- FastAPI
- MediaPipe
- YOLO
- ONNX
- Cloud Run
- Flutter
- MLflow
Public Signals
- API endpoints: 8 /, /app, /tma, /v1/*, /healthz
- Tasks: 3 pores, wrinkles, pores+wrinkles
- Image types: 5 jpeg/png/webp/tiff/bmp
- Default imgsz: 1280 segment endpoint default