CloudSoftwareHardware

ON THIS PAGE

  • Casting NN Blur
  • Demo
  • Source code
  • Pipeline

Casting NN Blur

This example demonstrates how to apply a blur effect using a neural network and the Cast node. The output of the cast node can be used in second stage neural networks or for further processing.

Demo

This example requires the DepthAI v3 API, see installation instructions.

Source code

Python
C++

Python

Python
GitHub
1#!/usr/bin/env python3
2
3import depthai as dai
4import cv2
5from pathlib import Path
6
7SHAPE = 300
8
9p = dai.Pipeline()
10
11camRgb = p.create(dai.node.ColorCamera)
12nn = p.create(dai.node.NeuralNetwork)
13rgbOut = p.create(dai.node.XLinkOut)
14cast = p.create(dai.node.Cast)
15castXout = p.create(dai.node.XLinkOut)
16
17camRgb.setPreviewSize(SHAPE, SHAPE)
18camRgb.setInterleaved(False)
19
20nnBlobPath = (Path(__file__).parent / Path('../models/blur_simplified_openvino_2021.4_6shave.blob')).resolve().absolute()
21
22nn.setBlobPath(nnBlobPath)
23
24rgbOut.setStreamName("rgb")
25
26castXout.setStreamName("cast")
27
28cast.setOutputFrameType(dai.RawImgFrame.Type.BGR888p)
29
30# Linking
31camRgb.preview.link(nn.input)
32camRgb.preview.link(rgbOut.input)
33nn.out.link(cast.input)
34cast.output.link(castXout.input)
35
36with dai.Device(p) as device:
37    qCam = device.getOutputQueue(name="rgb", maxSize=4, blocking=False)
38    qCast = device.getOutputQueue(name="cast", maxSize=4, blocking=False)
39
40
41    while True:
42        inCast = qCast.get()
43        assert isinstance(inCast, dai.ImgFrame)
44        inRgb = qCam.get()
45        assert isinstance(inRgb, dai.ImgFrame)
46        cv2.imshow("Blur", inCast.getCvFrame())
47        cv2.imshow("Original", inRgb.getCvFrame())
48
49
50        if cv2.waitKey(1) == ord('q'):
51            break

Pipeline

React Flow

Need assistance?

Head over to Discussion Forum for technical support or any other questions you might have.