A library that helps with embedding dynamic and generative images into Plotly Dash applications.

1. Install the dash-dynamic-images python package

pip install dash-dynamic-images

2. Add an html.Img element to your page's layout

This step does not differ in any way from the regular process of writing layouts for Dash applications.

import dash_html_components as html

...

app.layout = html.Div(children=[
    html.Img(id='image'),
    ...
    dcc.Input(id='x', type='number', value=10),
    dcc.Input(id='y', type='number', value=10)
])

3. Create an image_callback that serves your dynamic image

You can use image_callback decorator to create callbacks that return dynamic images. It works similarly to the standard Dash callback decorator, but with few notable differences:

• The first argument of the decorator should be an instance of the Dash object.
• The callback should have only one output, pointing at the src property of an Img layout element.
• The decorated function should return a PIL.Image.Image object (from the Pillow python library).

from dash_dynamic_images import image_callback
from PIL import Image, ImageDraw

...

@image_callback(
    app,
    dash.Output('image', 'src'),
    dash.Input('x', 'value'),
    dash.Input('y', 'value'))
def generate_image(x, y):
    image = Image.new('RGB', (200, 200), color=(0, 0, 200))
    ImageDraw.Draw(image).line([(0, 0), (x, y)], width=5)
    return image

As long as the returned object is a Pillow image, it does not matter on how was it create. You can generate it from scratch or obtain it from an external provider.

import requests

...

@image_callback(
    app,
    dash.Output('image', 'src'),
    dash.Input('button', 'n_clicks'))
def generate_image(_):
    response = requests.get('https://your_service.example/api/images/get')

    return Image.open(BytesIO(response.content))

Please consult the Pillow documentation for more details.

4. Enjoy the working application

A complete example of an application:

import dash
import dash_core_components as dcc
import dash_html_components as html
from dash_dynamic_images import image_callback
from PIL import Image, ImageDraw


app = dash.Dash()
app.layout = html.Div(children=[
    html.Img(id='image'),
    dcc.Input(id='x', type='number', value=10),
    dcc.Input(id='y', type='number', value=10)
])


@image_callback(
    app,
    dash.Output('image', 'src'),
    dash.Input('x', 'value'),
    dash.Input('y', 'value'))
def generate_image(x, y):
    image = Image.new('RGB', (200, 200), color=(0, 0, 200))
    ImageDraw.Draw(image).line([(0, 0), (x, y)], width=5)
    return image


if __name__ == '__main__':
    app.run_server(debug=True)

Whenever an image_callback is registered, the library performs two operations:

• It registers a flask route with a path of /image_generator/{unique_guid}.png that generates and serves images whenever invoked.
• It registers a standard Dash callback that produces and returns a parametrized (through the query string) image url based on the image_callback input values.

In practice, whenever one of the image_callback input parameters change, a new url is generated and inserted into the src property of the Img element, which in orders triggers a process of requesting and producing a new image.

The generated images are not persisted in the file system.

The library is aligned with the stateless nature of the Dash framework and therefor is compatible with its horizontal scaling capabilities (where a single application can be served by multiple processes and/or machines).

This library aims at simplifying the process described in the previous section so that it can be achieved through a single line of a python code.