#!/usr/bin/env python3

from pathlib import Path

import sys

import cv2

import depthai as dai

import numpy as np

# Get argument first

nnPath = str((Path(__file__).parent / Path('models/mobilenet-ssd_openvino_2021.2_6shave.blob')).resolve().absolute())

if len(sys.argv) > 1:

nnPath = sys.argv[1]

# Start defining a pipeline

pipeline = dai.Pipeline()

# Define a source - mono (grayscale) camera

camRight = pipeline.createMonoCamera()

camRight.setBoardSocket(dai.CameraBoardSocket.RIGHT)

camRight.setResolution(dai.MonoCameraProperties.SensorResolution.THE_720_P)

# Define a neural network that will make predictions based on the source frames

nn = pipeline.createMobileNetDetectionNetwork()

nn.setConfidenceThreshold(0.5)

nn.setBlobPath(nnPath)

nn.setNumInferenceThreads(2)

nn.input.setBlocking(False)

# Create a node to convert the grayscale frame into the nn-acceptable form

manip = pipeline.createImageManip()

manip.initialConfig.setResize(300, 300)

# The NN model expects BGR input. By default ImageManip output type would be same as input (gray in this case)

manip.initialConfig.setFrameType(dai.RawImgFrame.Type.BGR888p)

camRight.out.link(manip.inputImage)

manip.out.link(nn.input)

# Create outputs

manipOut = pipeline.createXLinkOut()

manipOut.setStreamName("right")

manip.out.link(manipOut.input)

nnOut = pipeline.createXLinkOut()

nnOut.setStreamName("nn")

nn.out.link(nnOut.input)

# MobilenetSSD label texts

labelMap = ["background", "aeroplane", "bicycle", "bird", "boat", "bottle", "bus", "car", "cat", "chair", "cow",

"diningtable", "dog", "horse", "motorbike", "person", "pottedplant", "sheep", "sofa", "train", "tvmonitor"]

# Pipeline defined, now the device is connected to

with dai.Device(pipeline) as device:

# Start pipeline

device.startPipeline()

# Output queues will be used to get the grayscale frames and nn data from the outputs defined above

qRight = device.getOutputQueue("right", maxSize=4, blocking=False)

qDet = device.getOutputQueue("nn", maxSize=4, blocking=False)

frame = None

detections = []

# nn data, being the bounding box locations, are in <0..1> range - they need to be normalized with frame width/height

def frameNorm(frame, bbox):

normVals = np.full(len(bbox), frame.shape[0])

normVals[::2] = frame.shape[1]

return (np.clip(np.array(bbox), 0, 1) * normVals).astype(int)

def displayFrame(name, frame):

for detection in detections:

bbox = frameNorm(frame, (detection.xmin, detection.ymin, detection.xmax, detection.ymax))

cv2.rectangle(frame, (bbox[0], bbox[1]), (bbox[2], bbox[3]), (255, 0, 0), 2)

cv2.putText(frame, labelMap[detection.label], (bbox[0] + 10, bbox[1] + 20), cv2.FONT_HERSHEY_TRIPLEX, 0.5, 255)

cv2.putText(frame, f"{int(detection.confidence * 100)}%", (bbox[0] + 10, bbox[1] + 40), cv2.FONT_HERSHEY_TRIPLEX, 0.5, 255)

cv2.imshow(name, frame)

while True:

# instead of get (blocking) used tryGet (nonblocking) which will return the available data or None otherwise

inRight = qRight.tryGet()

inDet = qDet.tryGet()

if inRight is not None:

frame = inRight.getCvFrame()

if inDet is not None:

detections = inDet.detections

if frame is not None:

displayFrame("right", frame)

if cv2.waitKey(1) == ord('q'):

break