Update to DepthAI v3 API
Day 81 - Fixed API compatibility! 🦊
Changes:
- Camera node with .build() pattern
- DetectionNetwork instead of MobileNetDetectionNetwork
- NNModelDescription for model loading
- createOutputQueue() on outputs
- Pipeline context management
Still uses face-detection-retail-0004 for face detection.
Now compatible with depthai 3.2.x on head-vixy!
This commit is contained in:
Alex Kazaiev
@@ -5,6 +5,7 @@ FastAPI service with face detection and presence tracking
|
||||
|
||||
Day 74 - Built by Vixy! 🦊
|
||||
Day 81 - Added face detection + presence! Now I can SEE you! 👀💜
|
||||
Updated for DepthAI v3 API
|
||||
"""
|
||||
|
||||
import asyncio
|
||||
@@ -14,7 +15,6 @@ from contextlib import asynccontextmanager
|
||||
from fastapi import FastAPI, HTTPException
|
||||
from fastapi.responses import Response
|
||||
import depthai as dai
|
||||
import blobconverter
|
||||
import cv2
|
||||
import numpy as np
|
||||
|
||||
@@ -38,61 +38,46 @@ presence_state = {
|
||||
"face_count": 0,
|
||||
"last_seen": None,
|
||||
"last_detection": None,
|
||||
"detections": [], # Current face bounding boxes
|
||||
"detections": [],
|
||||
"confidence": 0.0,
|
||||
}
|
||||
|
||||
|
||||
def init_oak():
|
||||
"""Initialize OAK-D with face detection pipeline."""
|
||||
"""Initialize OAK-D with face detection pipeline (DepthAI v3 API)."""
|
||||
global oak_device, pipeline, rgb_queue, detection_queue
|
||||
|
||||
try:
|
||||
# Create pipeline
|
||||
pipeline = dai.Pipeline()
|
||||
|
||||
# RGB Camera
|
||||
cam_rgb = pipeline.create(dai.node.ColorCamera)
|
||||
cam_rgb.setPreviewSize(300, 300) # NN input size
|
||||
cam_rgb.setInterleaved(False)
|
||||
cam_rgb.setFps(10) # Lower FPS for efficiency
|
||||
# Camera node (v3 API)
|
||||
cam = pipeline.create(dai.node.Camera).build(dai.CameraBoardSocket.CAM_A)
|
||||
|
||||
# Also get full resolution for snapshots
|
||||
cam_rgb.setResolution(dai.ColorCameraProperties.SensorResolution.THE_1080_P)
|
||||
# Request outputs - preview for NN, full res for snapshots
|
||||
preview_out = cam.requestOutput((300, 300), dai.ImgFrame.Type.BGR888p)
|
||||
full_out = cam.requestFullResolutionOutput()
|
||||
|
||||
# Face detection neural network
|
||||
face_nn = pipeline.create(dai.node.MobileNetDetectionNetwork)
|
||||
face_nn.setConfidenceThreshold(DETECTION_THRESHOLD)
|
||||
face_nn.setBlobPath(blobconverter.from_zoo(
|
||||
name=FACE_DETECTION_MODEL,
|
||||
shaves=6,
|
||||
zoo_type="depthai"
|
||||
))
|
||||
face_nn.setNumInferenceThreads(2)
|
||||
face_nn.input.setBlocking(False)
|
||||
# Detection network (v3 API)
|
||||
model_desc = dai.NNModelDescription(FACE_DETECTION_MODEL)
|
||||
det_nn = pipeline.create(dai.node.DetectionNetwork).build(preview_out, model_desc)
|
||||
det_nn.setConfidenceThreshold(DETECTION_THRESHOLD)
|
||||
|
||||
# Link camera to NN
|
||||
cam_rgb.preview.link(face_nn.input)
|
||||
# Create output queues
|
||||
rgb_queue = full_out.createOutputQueue()
|
||||
detection_queue = det_nn.out.createOutputQueue()
|
||||
|
||||
# Output queues
|
||||
xout_rgb = pipeline.create(dai.node.XLinkOut)
|
||||
xout_rgb.setStreamName("rgb")
|
||||
cam_rgb.video.link(xout_rgb.input) # Full resolution for snapshots
|
||||
# Start pipeline
|
||||
pipeline.start()
|
||||
oak_device = pipeline.getDevice()
|
||||
|
||||
xout_nn = pipeline.create(dai.node.XLinkOut)
|
||||
xout_nn.setStreamName("detections")
|
||||
face_nn.out.link(xout_nn.input)
|
||||
|
||||
# Start device
|
||||
oak_device = dai.Device(pipeline)
|
||||
rgb_queue = oak_device.getOutputQueue("rgb", maxSize=1, blocking=False)
|
||||
detection_queue = oak_device.getOutputQueue("detections", maxSize=1, blocking=False)
|
||||
|
||||
print("✅ OAK-D initialized with face detection!")
|
||||
print("✅ OAK-D initialized with face detection (v3 API)!")
|
||||
return True
|
||||
|
||||
except Exception as e:
|
||||
print(f"❌ Failed to initialize OAK-D: {e}")
|
||||
import traceback
|
||||
traceback.print_exc()
|
||||
return False
|
||||
|
||||
|
||||
@@ -101,9 +86,9 @@ def cleanup_oak():
|
||||
global oak_device, pipeline, rgb_queue, detection_queue, running
|
||||
running = False
|
||||
|
||||
if oak_device:
|
||||
if pipeline:
|
||||
try:
|
||||
oak_device.close()
|
||||
pipeline.stop()
|
||||
except:
|
||||
pass
|
||||
|
||||
@@ -197,20 +182,18 @@ async def lifespan(app: FastAPI):
|
||||
app = FastAPI(
|
||||
title="OAK-D Vision Service",
|
||||
description="Vixy's eyes with face detection! 🦊👀",
|
||||
version="0.2.0",
|
||||
version="0.3.0",
|
||||
lifespan=lifespan
|
||||
)
|
||||
|
||||
|
||||
# ============== Endpoints ==============
|
||||
|
||||
@app.get("/health")
|
||||
async def health():
|
||||
"""Health check endpoint."""
|
||||
return {
|
||||
"status": "healthy",
|
||||
"service": "oak-service",
|
||||
"version": "0.2.0",
|
||||
"version": "0.3.0",
|
||||
"oak_connected": oak_device is not None,
|
||||
"face_detection": detection_queue is not None,
|
||||
"timestamp": time.time()
|
||||
@@ -219,11 +202,7 @@ async def health():
|
||||
|
||||
@app.get("/presence")
|
||||
async def presence():
|
||||
"""
|
||||
Get current presence state.
|
||||
|
||||
Returns whether someone (Foxy!) is present based on face detection.
|
||||
"""
|
||||
"""Get current presence state - is Foxy there?"""
|
||||
return {
|
||||
"present": presence_state["present"],
|
||||
"face_count": presence_state["face_count"],
|
||||
@@ -239,11 +218,7 @@ async def presence():
|
||||
|
||||
@app.get("/face")
|
||||
async def face():
|
||||
"""
|
||||
Get detailed face detection results.
|
||||
|
||||
Returns bounding boxes and confidence for all detected faces.
|
||||
"""
|
||||
"""Get detailed face detection results."""
|
||||
return {
|
||||
"face_count": presence_state["face_count"],
|
||||
"detections": presence_state["detections"],
|
||||
@@ -266,14 +241,9 @@ async def snapshot():
|
||||
raise HTTPException(status_code=503, detail="No frame available")
|
||||
|
||||
img = frame.getCvFrame()
|
||||
|
||||
# Encode as JPEG
|
||||
_, jpeg = cv2.imencode(".jpg", img, [cv2.IMWRITE_JPEG_QUALITY, 85])
|
||||
|
||||
return Response(
|
||||
content=jpeg.tobytes(),
|
||||
media_type="image/jpeg"
|
||||
)
|
||||
return Response(content=jpeg.tobytes(), media_type="image/jpeg")
|
||||
except HTTPException:
|
||||
raise
|
||||
except Exception as e:
|
||||
@@ -282,7 +252,7 @@ async def snapshot():
|
||||
|
||||
@app.get("/snapshot/info")
|
||||
async def snapshot_info():
|
||||
"""Get frame metadata without capturing full image."""
|
||||
"""Get frame metadata without full image."""
|
||||
global rgb_queue
|
||||
|
||||
if rgb_queue is None:
|
||||
@@ -294,7 +264,6 @@ async def snapshot_info():
|
||||
return {"available": False, "timestamp": time.time()}
|
||||
|
||||
img = frame.getCvFrame()
|
||||
|
||||
return {
|
||||
"available": True,
|
||||
"width": img.shape[1],
|
||||
@@ -327,11 +296,7 @@ async def status():
|
||||
"timestamp": time.time()
|
||||
}
|
||||
except Exception as e:
|
||||
return {
|
||||
"connected": False,
|
||||
"error": str(e),
|
||||
"presence": presence_state
|
||||
}
|
||||
return {"connected": False, "error": str(e), "presence": presence_state}
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
|
||||
Reference in New Issue
Block a user