""" Face Inference Microservice for Hugging Face Spaces Extracts face embeddings and matches against known encodings """ from fastapi import FastAPI, HTTPException, UploadFile, File, Form from fastapi.responses import JSONResponse, HTMLResponse, FileResponse from fastapi.staticfiles import StaticFiles import asyncio import numpy as np import io import base64 from PIL import Image import cv2 import logging from insightface.app import FaceAnalysis import os logging.basicConfig(level=logging.INFO) logger = logging.getLogger(__name__) # Global face recognition model (lazy-loaded on first use) face_app = None face_app_lock = asyncio.Lock() async def get_face_app(): global face_app if face_app is not None: return face_app async with face_app_lock: if face_app is not None: return face_app try: logger.info("Loading face recognition model...") model_name = os.environ.get("FACE_RECOGNITION_MODEL", "buffalo_l") app_instance = FaceAnalysis(name=model_name, providers=["CPUExecutionProvider"]) app_instance.prepare(ctx_id=-1) face_app = app_instance logger.info("✓ Model loaded successfully") except Exception as e: logger.error(f"Failed to load model: {e}") raise return face_app app = FastAPI(title="Face Inference Service", version="1.0") # Mount static files static_dir = os.path.join(os.path.dirname(__file__), "static") if os.path.exists(static_dir): app.mount("/static", StaticFiles(directory=static_dir), name="static") logger.info(f"✓ Static files mounted from {static_dir}") @app.get("/", response_class=HTMLResponse) async def root(): """Serve the web UI""" html_path = os.path.join(os.path.dirname(__file__), "static", "index.html") if os.path.exists(html_path): with open(html_path, "r", encoding="utf-8") as f: return f.read() return "

Face Inference Service

API is running. Visit /docs for API documentation.

" @app.get("/health") async def health_check(): """Health check endpoint""" return { "status": "healthy", "service": "face-inference", "model_loaded": face_app is not None, } @app.post("/recognize") async def recognize_face( frame_data: str = Form(...), known_encodings: str = Form(...), tolerance: float = Form(default=0.85) ): """ Recognize faces in a frame against known encodings Args: frame_data: Base64 encoded image known_encodings: JSON string of numpy array (serialized) tolerance: Distance threshold for matching Returns: List of matched student IDs and distances """ try: face_app_instance = await get_face_app() # Decode frame img_bytes = base64.b64decode(frame_data.split(',')[1] if ',' in frame_data else frame_data) img = Image.open(io.BytesIO(img_bytes)).convert('RGB') frame = np.array(img) # Detect faces faces = face_app_instance.get(frame) if not faces: return {"success": True, "recognized": [], "count": 0} # Parse known encodings import json known_data = json.loads(known_encodings) known_encodings_arr = np.array(known_data['encodings']) known_metadata = known_data.get('metadata', []) recognized = [] # Match each detected face for face in faces: embedding = face.normed_embedding distances = np.linalg.norm(known_encodings_arr - embedding, axis=1) min_idx = np.argmin(distances) min_distance = float(distances[min_idx]) if min_distance < tolerance: student_info = known_metadata[min_idx] recognized.append({ "roll_no": student_info.get('roll_no', ''), "name": student_info.get('name', ''), "distance": min_distance }) logger.info(f"✓ Recognized: {student_info.get('name')} - Distance: {min_distance:.3f}") return { "success": True, "recognized": recognized, "count": len(recognized), "faces_detected": len(faces) } except json.JSONDecodeError as e: logger.error(f"JSON parse error: {e}") raise HTTPException(status_code=400, detail="Invalid encoding data format") except Exception as e: logger.error(f"Recognition error: {e}") raise HTTPException(status_code=500, detail=str(e)) @app.post("/recognize_batch") async def recognize_batch( frames: list = Form(...), known_encodings: str = Form(...), tolerance: float = Form(default=0.85) ): """ Batch recognition for multiple frames """ try: face_app_instance = await get_face_app() import json known_data = json.loads(known_encodings) known_encodings_arr = np.array(known_data['encodings']) known_metadata = known_data.get('metadata', []) all_recognized = set() for frame_data in frames: try: img_bytes = base64.b64decode(frame_data.split(',')[1] if ',' in frame_data else frame_data) img = Image.open(io.BytesIO(img_bytes)).convert('RGB') frame = np.array(img) faces = face_app_instance.get(frame) for face in faces: embedding = face.normed_embedding distances = np.linalg.norm(known_encodings_arr - embedding, axis=1) min_idx = np.argmin(distances) min_distance = float(distances[min_idx]) if min_distance < tolerance: student_info = known_metadata[min_idx] all_recognized.add(student_info.get('roll_no', '')) except Exception as e: logger.warning(f"Error processing frame in batch: {e}") continue return { "success": True, "recognized": list(all_recognized), "count": len(all_recognized) } except Exception as e: logger.error(f"Batch recognition error: {e}") raise HTTPException(status_code=500, detail=str(e)) @app.post("/get_embedding") async def get_embedding( image_data: str = Form(...) ): """ Extract embedding from a single image """ try: face_app_instance = await get_face_app() img_bytes = base64.b64decode(image_data.split(',')[1] if ',' in image_data else image_data) img = Image.open(io.BytesIO(img_bytes)).convert('RGB') frame = np.array(img) faces = face_app_instance.get(frame) if not faces: raise HTTPException(status_code=400, detail="No face detected in image") embedding = faces[0].normed_embedding.tolist() return { "success": True, "embedding": embedding, "faces_detected": len(faces) } except Exception as e: logger.error(f"Embedding error: {e}") raise HTTPException(status_code=500, detail=str(e)) if __name__ == "__main__": import uvicorn port = int(os.environ.get("PORT", 7860)) uvicorn.run(app, host="0.0.0.0", port=port)