vixy/oak-service
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

facial recognition

Browse Source
This commit is contained in:
Alex
2026-02-01 11:36:22 -06:00
parent 3c273d7d02
commit 3ac8778cac
3 changed files with 507 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

177
oak_service_spatial.py
View File

@@ -11,13 +11,20 @@ Day 82 - SPATIAL UPGRADE! Now I know how far away you are! 📏🦊
import time
import threading
import logging
from pathlib import Path
from contextlib import asynccontextmanager
from fastapi import FastAPI, HTTPException
from fastapi import FastAPI, File, Form, HTTPException, UploadFile
from fastapi.responses import Response
import depthai as dai
import cv2
import numpy as np
from face_recognition import FaceRecognizer
logger = logging.getLogger("oak-service")
logging.basicConfig(level=logging.INFO)
# ============== Configuration ==============
DETECTION_MODEL = "yolov6-nano" # Has 'person' class
PERSON_CLASS_ID = 0 # 'person' is class 0 in COCO
@@ -29,6 +36,12 @@ DETECTION_INTERVAL = 0.5
DEPTH_LOWER_THRESHOLD = 100 # 10cm minimum
DEPTH_UPPER_THRESHOLD = 10000 # 10m maximum
# Face recognition models
MODELS_DIR = Path(__file__).parent / "models"
FACE_DETECT_MODEL = MODELS_DIR / "ssd_mobilenet_v2_face_quant_postprocess_edgetpu.tflite"
FACE_EMBED_MODEL = MODELS_DIR / "facenet.tflite"
FACE_DB_PATH = Path(__file__).parent / "faces.db"
# ============== Global State ==============
pipeline_ctx = None
detection_queue = None
@@ -37,6 +50,7 @@ depth_queue = None
detection_thread = None
running = False
labels = []
face_recognizer = None
presence_state = {
"present": False,
@@ -45,14 +59,35 @@ presence_state = {
"last_detection": None,
"detections": [],
"confidence": 0.0,
# NEW: spatial data!
# Spatial data
"distance_mm": None,
"spatial_x": None,
"spatial_y": None,
"spatial_z": None,
# Face recognition
"recognized_name": None,
"recognition_confidence": None,
}
def init_face_recognition():
"""Initialize Coral face detection + FaceNet embedding."""
global face_recognizer
try:
face_recognizer = FaceRecognizer(
face_model_path=FACE_DETECT_MODEL,
embed_model_path=FACE_EMBED_MODEL,
db_path=FACE_DB_PATH,
)
print("✅ Face recognition initialized (Coral + FaceNet)")
return True
except Exception as e:
print(f"⚠️ Face recognition unavailable: {e}")
import traceback
traceback.print_exc()
return False
def init_oak():
"""Initialize OAK-D with SPATIAL person detection pipeline (depthai v3)."""
global pipeline_ctx, detection_queue, rgb_queue, depth_queue, labels
@@ -123,9 +158,13 @@ def init_oak():
def cleanup_oak():
"""Cleanup OAK-D resources."""
global pipeline_ctx, running
global pipeline_ctx, running, face_recognizer
running = False
if face_recognizer:
face_recognizer.close()
face_recognizer = None
if pipeline_ctx:
try:
pipeline_ctx.stop()
@@ -162,30 +201,60 @@ def detection_loop():
if person_count > 0:
presence_state["present"] = True
presence_state["last_seen"] = now
# Get highest confidence detection
best = max(persons, key=lambda d: d.confidence)
presence_state["confidence"] = best.confidence
# SPATIAL DATA! 🎉
# Spatial data
presence_state["spatial_x"] = best.spatialCoordinates.x
presence_state["spatial_y"] = best.spatialCoordinates.y
presence_state["spatial_z"] = best.spatialCoordinates.z
presence_state["distance_mm"] = best.spatialCoordinates.z # Z is depth
presence_state["detections"] = [
{
presence_state["distance_mm"] = best.spatialCoordinates.z
# Face recognition
face_results = []
if face_recognizer and rgb_queue:
rgb_data = rgb_queue.tryGet()
if rgb_data is not None:
rgb_frame = rgb_data.getCvFrame()
try:
face_results = face_recognizer.process_frame(
rgb_frame, persons
)
except Exception as e:
logger.warning("Face recognition error: %s", e)
det_list = []
best_recognized = None
best_recog_conf = 0.0
for i, d in enumerate(persons):
det = {
"xmin": d.xmin, "ymin": d.ymin,
"xmax": d.xmax, "ymax": d.ymax,
"confidence": d.confidence,
# Spatial coordinates in mm
"x_mm": d.spatialCoordinates.x,
"y_mm": d.spatialCoordinates.y,
"z_mm": d.spatialCoordinates.z,
"distance_m": d.spatialCoordinates.z / 1000.0,
"recognized_name": None,
"recognition_confidence": None,
}
for d in persons
]
if i < len(face_results):
det["recognized_name"] = face_results[i]["recognized_name"]
det["recognition_confidence"] = face_results[i]["recognition_confidence"]
if det["recognized_name"] and (
det["recognition_confidence"] or 0
) > best_recog_conf:
best_recognized = det["recognized_name"]
best_recog_conf = det["recognition_confidence"]
det_list.append(det)
presence_state["detections"] = det_list
presence_state["recognized_name"] = best_recognized
presence_state["recognition_confidence"] = (
round(best_recog_conf, 3) if best_recognized else None
)
else:
presence_state["detections"] = []
presence_state["confidence"] = 0.0
@@ -193,7 +262,9 @@ def detection_loop():
presence_state["spatial_y"] = None
presence_state["spatial_z"] = None
presence_state["distance_mm"] = None
presence_state["recognized_name"] = None
presence_state["recognition_confidence"] = None
# Check timeout
if presence_state["last_seen"]:
if now - presence_state["last_seen"] > PRESENCE_TIMEOUT:
@@ -214,7 +285,9 @@ async def lifespan(app: FastAPI):
global running, detection_thread
print("🦊 Starting OAK-D SPATIAL Vision Service...")
init_face_recognition()
if init_oak():
running = True
detection_thread = threading.Thread(target=detection_loop, daemon=True)
@@ -231,8 +304,8 @@ async def lifespan(app: FastAPI):
app = FastAPI(
title="OAK-D SPATIAL Vision Service",
description="Vixy's eyes with SPATIAL presence detection! 🦊👀📏",
version="0.4.0",
description="Vixy's eyes with SPATIAL presence detection + face recognition! 🦊👀📏",
version="0.5.0",
lifespan=lifespan
)
@@ -243,10 +316,11 @@ async def health():
return {
"status": "healthy",
"service": "oak-service",
"version": "0.4.0",
"version": "0.5.0",
"oak_connected": pipeline_ctx is not None,
"detection_model": DETECTION_MODEL,
"spatial_enabled": True,
"face_recognition_enabled": face_recognizer is not None,
"timestamp": time.time()
}
@@ -267,7 +341,6 @@ async def presence():
if presence_state["last_seen"] else None
),
"confidence": presence_state["confidence"],
# SPATIAL DATA
"distance_mm": presence_state["distance_mm"],
"distance_m": distance_m,
"spatial": {
@@ -275,6 +348,8 @@ async def presence():
"y_mm": presence_state["spatial_y"],
"z_mm": presence_state["spatial_z"],
} if presence_state["spatial_z"] else None,
"recognized_name": presence_state["recognized_name"],
"recognition_confidence": presence_state["recognition_confidence"],
"timestamp": time.time()
}
@@ -340,6 +415,68 @@ async def depth_frame():
raise HTTPException(status_code=500, detail=str(e))
# ============== Face Enrollment API ==============
@app.post("/faces/enroll")
async def enroll_face_upload(
name: str = Form(...),
photo: UploadFile = File(...),
):
"""Enroll a face by uploading a photo (multipart form: name + photo)."""
if face_recognizer is None:
raise HTTPException(status_code=503, detail="Face recognition not available")
contents = await photo.read()
nparr = np.frombuffer(contents, np.uint8)
image = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
if image is None:
raise HTTPException(status_code=400, detail="Could not decode image")
result = face_recognizer.enroll(name, image)
if not result["success"]:
raise HTTPException(status_code=400, detail=result["error"])
return result
@app.post("/faces/enroll-from-camera")
async def enroll_face_camera(name: str):
"""Enroll a face using the current camera frame. Pass name as query param."""
if face_recognizer is None:
raise HTTPException(status_code=503, detail="Face recognition not available")
if rgb_queue is None:
raise HTTPException(status_code=503, detail="Camera not available")
frame_data = rgb_queue.tryGet()
if frame_data is None:
raise HTTPException(status_code=503, detail="No frame available")
image = frame_data.getCvFrame()
result = face_recognizer.enroll(name, image)
if not result["success"]:
raise HTTPException(status_code=400, detail=result["error"])
return result
@app.get("/faces")
async def list_faces():
"""List enrolled faces."""
if face_recognizer is None:
raise HTTPException(status_code=503, detail="Face recognition not available")
return {"faces": face_recognizer.list_faces()}
@app.delete("/faces/{name}")
async def delete_face(name: str):
"""Remove all embeddings for a person."""
if face_recognizer is None:
raise HTTPException(status_code=503, detail="Face recognition not available")
result = face_recognizer.delete_face(name)
if not result["success"]:
raise HTTPException(status_code=404, detail=f"No face found for '{name}'")
return result
if __name__ == "__main__":
import uvicorn
uvicorn.run(app, host="0.0.0.0", port=8100)