Hello World

Learn how to use the DepthAI Python API to display a color video stream.



Let’s get your development environment setup first. This tutorial uses:

  • Python >=3.6

  • DepthAI Python API

  • cv2, blobconverter and numpy Python modules.

Code Overview

The depthai Python module provides access to your board’s 4K 60 Hz color camera. We’ll display a video stream from this camera to your desktop. You can find the complete source code for this tutorial on GitHub.

File Setup

Setup the following file structure on your computer:

cd ~
mkdir -p depthai-tutorials-practice/1-hello-world
touch depthai-tutorials-practice/1-hello-world/hello_world.py
cd depthai-tutorials-practice/1-hello-world

What’s with the -practice suffix in parent directory name? Our tutorials are available on GitHub via the depthai-tutorials repository. We’re appending -practice so you can distinguish between your work and our finished tutorials (should you choose to download those).

Install pip dependencies

To display the DepthAI color video stream we need to import a small number of packages. Download and install the requirements for this tutorial:

python3 -m pip install numpy opencv-python depthai blobconverter --user

Test your environment

Let’s verify we’re able to load all of our dependencies. Open the hello_world.py file you created earlier in your code editor. Copy and paste the following into hello_world.py:

import numpy as np  # numpy - manipulate the packet data returned by depthai
import cv2  # opencv - display the video stream
import depthai  # depthai - access the camera and its data packets
import blobconverter  # blobconverter - compile and download MyriadX neural network blobs

Try running the script and ensure it executes without error:

python3 hello_world.py

If you see the following error:

ModuleNotFoundError: No module named 'depthai'

…follow these steps in our troubleshooting section.

Define a pipeline

Any action from DepthAI, whether it’s a neural inference or color camera output, require a pipeline to be defined, including nodes and connections corresponding to our needs.

In this case, we want to see the frames from color camera, as well as a simple neural network to be ran on top of them.

Let’s start off with an empty Pipeline object

pipeline = depthai.Pipeline()

Now, first node we will add is a ColorCamera. We will use the preview output, resized to 300x300 to fit the mobilenet-ssd input size (which we will define later)

cam_rgb = pipeline.create(depthai.node.ColorCamera)
cam_rgb.setPreviewSize(300, 300)

Up next, let’s define a MobileNetDetectionNetwork node with mobilenet-ssd network. The blob file for this example will be compiled and downloaded automatically using blobconverter tool. blobconverter.from_zoo() function returns Path to the model, so we can directly put it inside the detection_nn.setBlobPath() function. With this node, the output from nn will be parsed on device side and we’ll receive a ready to use detection objects. For this to work properly, we need also to set the confidence threshold to filter out the incorrect results

detection_nn = pipeline.create(depthai.node.MobileNetDetectionNetwork)
# Set path of the blob (NN model). We will use blobconverter to convert&download the model
# detection_nn.setBlobPath("/path/to/model.blob")
detection_nn.setBlobPath(blobconverter.from_zoo(name='mobilenet-ssd', shaves=6))

And now, let’s connect a color camera preview output to neural network input


Finally, we want to receive both color camera frames and neural network inference results - as these are produced on the device, they need to be transported to our machine (host). The communication between device and host is handled by XLink, and in our case, since we want to receive data from device to host, we will use XLinkOut node

xout_rgb = pipeline.create(depthai.node.XLinkOut)

xout_nn = pipeline.create(depthai.node.XLinkOut)

Initialize the DepthAI Device

Having the pipeline defined, we can now initialize a device with pipeline and start it

with depthai.Device(pipeline) as device:


By default, the DepthAI is accessed as a USB3 device. This comes with several limitations.

If you’d like to communicate via USB2, being free from these but having a limited bandwidth, initialize the DepthAI with the following code

device = depthai.Device(pipeline, usb2Mode=True)

From this point on, the pipeline will be running on the device, producing results we requested. Let’s grab them

Adding helpers

As XLinkOut nodes has been defined in the pipeline, we’ll define now a host side output queues to access the produced results

q_rgb = device.getOutputQueue("rgb")
q_nn = device.getOutputQueue("nn")

These will fill up with results, so next thing to do is consume the results. We will need two placeholders - one for rgb frame and one for nn results

frame = None
detections = []

Also, due to neural network implementation details, bounding box coordinates in inference results are represented as floats from <0..1> range - so relative to frame width/height (e.g. if image has 200px width and nn returned x_min coordinate equal to 0.2, this means the actual (normalised) x_min coordinate is 40px).

That’s why we need to define a helper function, frameNorm, that will convert these <0..1> values into actual pixel positions

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)

Consuming the results

Having everything prepared, we are ready to start out main program loop

while True:
  # ...

Now, inside this loop, first thing to do is fetching latest results from both nn node and color camera

in_rgb = q_rgb.tryGet()
in_nn = q_nn.tryGet()

The tryGet method returns either the latest result or None if the queue is empty.

Results, both from rgb camera or neural network, will be delivered as 1D arrays, so both of them will require transformations to be useful for display (we have already defined one of the transformations needed - the frameNorm function)

First up, if we receive a frame from rgb camera using the getCvFrame command

if in_rgb is not None:
    frame = in_rgb.getCvFrame()

Second, we will receive the neural network results. Default MobileNetSSD result has 7 fields, each being respectively image_id, label, confidence, x_min, y_min, x_max, y_max, and by accessing the detections array, we receive the detection objects that allow us to access these fields

if in_nn is not None:
    detections = in_nn.detections

Display the results

Up to this point, we have all our results consumed from the DepthaI device, and only thing left is to actually display them.

if frame is not None:
    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.imshow("preview", frame)

You can see here the usage of frameNorm we defined earlier for bounding box coordinates normalization. By using cv2.rectangle we draw a rectangle on the rgb frame as an indicator where the face position is, and then we display the frame using cv2.imshow

Finally, we add a way to terminate our program (as it’s running inside an infinite loop). We will use cv2.waitKey method, that waits for a key to be pressed by user - in our case, we want to break out of the loop when user presses q key

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

Running the example

Putting it all together, only thing left to do is to run the file we’ve prepared in this tutorial and see the results

python3 hello_world.py

You’re on your way! You can find the complete code for this tutorial on GitHub.

Got questions?

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