🔎 SafeFind — Missing Persons Recovery Platform (AI Age Progression + Face Recognition)

A production-ready backend that combines age progression and face recognition to locate missing people.
Built with Flask (Python) and Firebase (Auth, Firestore, Storage), with an admin surface for moderating reports and users.

---

✨ Key Features

• Account & Profiles

  • Register / Login (Firebase Auth)
  • Update profile, avatar, phone
  • Role-based access: user, admin

• Posts (Missing / Found)

  • Create/update/delete posts with images and metadata
  • Report suspicious/spam posts
  • List & view posts (recent first)

• Search

  • Upload a photo → system age-progresses the face to a target age → runs face similarity against database → returns the closest match (with uploader contact, if available)

• Admin Tools

  • List users (paged), suspend/unsuspend users
  • Review post reports, delete posts
  • Stats: users, found posts, reports, successful / unsuccessful matches

• Image Pipeline

  • Type & size validation, blur detection, square crop, resize, JPEG encode
  • Public URL upload to Firebase Storage

---

🧱 Tech Stack

• Backend: Flask (Python 3.11+)
• AI / CV: Age progression service (FastAPI endpoint), FaceNet embeddings + cosine distance
• Data: Firebase Firestore + Storage
• Auth: Firebase Auth (ID tokens)
• Image: OpenCV, PIL
• ML tooling: PyTorch (age progression), LFW/FFHQ datasets (training utilities)

---

🗺️ Architecture (high level)

flowchart LR
  Mobile[Flutter App] --> API[Flask API]
  Admin[Admin Panel] --> API
  API --> Auth[Firebase Auth]
  API --> Store[Firestore]
  API --> Files[Firebase Storage]
  API --> Age[Age Progression Service]
  API --> Face[Face Recognition]

Loading

---

🧩 API (summary)

Auth

• POST /register — create account (validates email/phone/password)
• POST /login — password login (returns tokens + profile)
• PATCH /update-profile — update profile (auth)
• GET /me — current user (auth)

Posts

• POST /posts/missing — create missing-person post (auth; multipart: image_file + fields)
• POST /posts/found — create found-person post (auth; multipart)
• PATCH /posts/<post_id> — update post (auth; can replace image)
• DELETE /posts/<post_id> — delete own post (auth)
• GET /posts — list posts (recent first)
• GET /posts/<post_id> — get a post
• POST /posts/<post_id>/report — report a post (auth)

Search & Age Progression

• POST /age-progress — multipart (image, target_age) or JSON (image_b64, target_age); returns progressed image URL and closest match (if any)
• POST /search — image-based search across candidates; returns best match with distance

Admin (prefix /admin, admin-only)

• GET /users — list users (paged)
• PATCH /users/<uid>/status — suspend/unsuspend user
• GET /reports — list post reports
• DELETE /posts/<post_id> — delete any post
• GET /matches/successful/count — successful matches count
• GET /matches/unsuccessful/count — unsuccessful matches count

---

🗃️ Data Model (core)

User

• uid, email, first_name, last_name, phone, photo_url, role("user"|"admin"), created_at
• Helpers: is_owner(uid), get_full_name(), get_created_at_iso()

Post

• id, uid, author_name, post_type("missing"|"found"), image_url, created_at, status
• payload (typed by post_type), e.g.
  • missing: missing_name, missing_age, last_seen, notes, gender, ...
  • found: found_name, estimated_age, found_location, gender, ...

---

🖼️ Image Preprocessing

• Accepts JPEG/PNG
• Rejects small images (min(side) < MIN_IMAGE_SIDE) or blurry images (Laplacian variance < threshold)
• Center crop to square, resize (≤ IMAGE_RESIZE_TO), encode JPEG (JPEG_QUALITY)
• Uploads to Storage under:
  • missing_posts/{uid}/{uuid}.jpg
  • found_posts/{uid}/{uuid}.jpg

---

🧠 Face Recognition & Age Progression

• Face recognition: FaceNet embeddings + cosine distance; lower is more similar; default threshold 0.40
• Age progression: external FastAPI endpoint (/age-transform) that returns a transformed face; backend uploads the result then performs similarity search

---

🛡️ Security & Privacy

• Auth: Firebase ID tokens on all user actions; admin_required guard for admin routes
• PII: Stores minimal necessary fields; images stored on Firebase Storage
• Abuse handling: Post reporting + admin moderation
• Transport: Use HTTPS for all public endpoints; secure service account JSON

---

🧪 (Optional) Training Utilities

• Age progression model (prototype): PyTorch encoder–age–decoder trained on FFHQ; Google Drive download helpers; saves FFHQ_age_progression.pth
• Face recognition pairs: Utility to create LFW positive/negative pairs for Siamese training/eval

---

✅ Checklist for Deployment

• Set all environment variables on your host
• Upload service account JSON and set GOOGLE_APPLICATION_CREDENTIALS
• Create Firestore indexes if needed (composite by created_at)
• Lock down Storage rules appropriately
• Point AGE_API_BASE_URL to your FastAPI/Colab/ngrok endpoint

---