26.2 - Mono & MobilenetSSD with spatial data

This example shows how to run MobileNetv2SSD on the rectified right input frame, and how to display both the preview, detections, depth map and spatial information (X,Y,Z). It’s similar to example ‘21_mobilenet_decoding_on_device’ except it has spatial data. X,Y,Z coordinates are relative to the center of depth map.

setConfidenceThreshold - confidence threshold above which objects are detected

Demo

Setup

Please run the following command to install the required dependencies

 python3 -m pip install -U pip
 python3 -m pip install opencv-python
 python3 -m pip install -U --force-reinstall depthai

For additional information, please follow installation guide

This example also requires MobilenetSDD blob (mobilenet-ssd_openvino_2021.2_6shave.blob file) to work - you can download it from here

Source code

Also available on GitHub

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
#!/usr/bin/env python3

from pathlib import Path
import sys
import cv2
import depthai as dai
import numpy as np
import time

'''
Mobilenet SSD device side decoding demo
  The "mobilenet-ssd" model is a Single-Shot multibox Detection (SSD) network intended
  to perform object detection. This model is implemented using the Caffe* framework.
  For details about this model, check out the repository <https://github.com/chuanqi305/MobileNet-SSD>.
'''

# 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"]

syncNN = True
flipRectified = True

# 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]

if not Path(nnPath).exists():
    import sys
    raise FileNotFoundError(f'Required file/s not found, please run "{sys.executable} install_requirements.py"')

# Start defining a pipeline
pipeline = dai.Pipeline()


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.ImgFrame.Type.BGR888p)
# manip.setKeepAspectRatio(False)

# Define a neural network that will make predictions based on the source frames
spatialDetectionNetwork = pipeline.createMobileNetSpatialDetectionNetwork()
spatialDetectionNetwork.setConfidenceThreshold(0.5)
spatialDetectionNetwork.setBlobPath(nnPath)
spatialDetectionNetwork.input.setBlocking(False)
spatialDetectionNetwork.setBoundingBoxScaleFactor(0.5)
spatialDetectionNetwork.setDepthLowerThreshold(100)
spatialDetectionNetwork.setDepthUpperThreshold(5000)

manip.out.link(spatialDetectionNetwork.input)

# Create outputs
xoutManip = pipeline.createXLinkOut()
xoutManip.setStreamName("right")
if syncNN:
    spatialDetectionNetwork.passthrough.link(xoutManip.input)
else:
    manip.out.link(xoutManip.input)

depthRoiMap = pipeline.createXLinkOut()
depthRoiMap.setStreamName("boundingBoxDepthMapping")

xoutDepth = pipeline.createXLinkOut()
xoutDepth.setStreamName("depth")

nnOut = pipeline.createXLinkOut()
nnOut.setStreamName("detections")
spatialDetectionNetwork.out.link(nnOut.input)
spatialDetectionNetwork.boundingBoxMapping.link(depthRoiMap.input)

monoLeft = pipeline.createMonoCamera()
monoRight = pipeline.createMonoCamera()
stereo = pipeline.createStereoDepth()
monoLeft.setResolution(dai.MonoCameraProperties.SensorResolution.THE_400_P)
monoLeft.setBoardSocket(dai.CameraBoardSocket.LEFT)
monoRight.setResolution(dai.MonoCameraProperties.SensorResolution.THE_400_P)
monoRight.setBoardSocket(dai.CameraBoardSocket.RIGHT)
stereo.setConfidenceThreshold(255)

stereo.rectifiedRight.link(manip.inputImage)

monoLeft.out.link(stereo.left)
monoRight.out.link(stereo.right)

stereo.depth.link(spatialDetectionNetwork.inputDepth)
spatialDetectionNetwork.passthroughDepth.link(xoutDepth.input)

# Connect and start the pipeline
with dai.Device(pipeline) as device:

    # Output queues will be used to get the rgb frames and nn data from the outputs defined above
    previewQueue = device.getOutputQueue(name="right", maxSize=4, blocking=False)
    detectionNNQueue = device.getOutputQueue(name="detections", maxSize=4, blocking=False)
    depthRoiMap = device.getOutputQueue(name="boundingBoxDepthMapping", maxSize=4, blocking=False)
    depthQueue = device.getOutputQueue(name="depth", maxSize=4, blocking=False)

    rectifiedRight = None
    detections = []

    startTime = time.monotonic()
    counter = 0
    fps = 0
    color = (255, 255, 255)

    while True:
        inRectified = previewQueue.get()
        det = detectionNNQueue.get()
        depth = depthQueue.get()

        counter += 1
        currentTime = time.monotonic()
        if (currentTime - startTime) > 1:
            fps = counter / (currentTime - startTime)
            counter = 0
            startTime = currentTime

        rectifiedRight = inRectified.getCvFrame()

        depthFrame = depth.getFrame()

        depthFrameColor = cv2.normalize(depthFrame, None, 255, 0, cv2.NORM_INF, cv2.CV_8UC1)
        depthFrameColor = cv2.equalizeHist(depthFrameColor)
        depthFrameColor = cv2.applyColorMap(depthFrameColor, cv2.COLORMAP_HOT)
        detections = det.detections
        if len(detections) != 0:
            boundingBoxMapping = depthRoiMap.get()
            roiDatas = boundingBoxMapping.getConfigData()

            for roiData in roiDatas:
                roi = roiData.roi
                roi = roi.denormalize(depthFrameColor.shape[1], depthFrameColor.shape[0])
                topLeft = roi.topLeft()
                bottomRight = roi.bottomRight()
                xmin = int(topLeft.x)
                ymin = int(topLeft.y)
                xmax = int(bottomRight.x)
                ymax = int(bottomRight.y)
                cv2.rectangle(depthFrameColor, (xmin, ymin), (xmax, ymax), color, cv2.FONT_HERSHEY_SCRIPT_SIMPLEX)

        if flipRectified:
            rectifiedRight = cv2.flip(rectifiedRight, 1)

        # If the rectifiedRight is available, draw bounding boxes on it and show the rectifiedRight
        height = rectifiedRight.shape[0]
        width = rectifiedRight.shape[1]
        for detection in detections:
            if flipRectified:
                swap = detection.xmin
                detection.xmin = 1 - detection.xmax
                detection.xmax = 1 - swap
            # Denormalize bounding box
            x1 = int(detection.xmin * width)
            x2 = int(detection.xmax * width)
            y1 = int(detection.ymin * height)
            y2 = int(detection.ymax * height)

            try:
                label = labelMap[detection.label]
            except:
                label = detection.label

            cv2.putText(rectifiedRight, str(label), (x1 + 10, y1 + 20), cv2.FONT_HERSHEY_TRIPLEX, 0.5, color)
            cv2.putText(rectifiedRight, "{:.2f}".format(detection.confidence*100), (x1 + 10, y1 + 35), cv2.FONT_HERSHEY_TRIPLEX, 0.5, color)
            cv2.putText(rectifiedRight, f"X: {int(detection.spatialCoordinates.x)} mm", (x1 + 10, y1 + 50), cv2.FONT_HERSHEY_TRIPLEX, 0.5, color)
            cv2.putText(rectifiedRight, f"Y: {int(detection.spatialCoordinates.y)} mm", (x1 + 10, y1 + 65), cv2.FONT_HERSHEY_TRIPLEX, 0.5, color)
            cv2.putText(rectifiedRight, f"Z: {int(detection.spatialCoordinates.z)} mm", (x1 + 10, y1 + 80), cv2.FONT_HERSHEY_TRIPLEX, 0.5, color)

            cv2.rectangle(rectifiedRight, (x1, y1), (x2, y2), color, cv2.FONT_HERSHEY_SIMPLEX)

        cv2.putText(rectifiedRight, "NN fps: {:.2f}".format(fps), (2, rectifiedRight.shape[0] - 4), cv2.FONT_HERSHEY_TRIPLEX, 0.4, color)
        cv2.imshow("depth", depthFrameColor)
        cv2.imshow("rectified right", rectifiedRight)

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

Got questions?

We’re always happy to help with code or other questions you might have.