Jupyter Splitview

A cell magic that displays images in splitview.

Installation

pip install jupyter-splitview

Example

import jupyter_splitview

%%splity

from skimage import data
from skimage.color import rgb2gray
import matplotlib.pyplot as plt

img = data.chelsea()
grayscale_img = rgb2gray(img)

fig, ax1 = plt.subplots()
ax1.axis("off")
ax1.imshow(img)

fig, ax2 = plt.subplots()
ax2.axis("off")
ax2.imshow(grayscale_img, cmap="gray")

Note: The split view widget is still responsive after closing and reopening the notebook without running the cell again.

Another example:

%%splity --position 73% --height auto

import matplotlib.pyplot as plt
import numpy as np

array1 = np.full((15, 30), 10)
array2 = np.random.randint(0, 10, size=(15, 30))
fig, ax1 = plt.subplots(figsize=(5, 10))
ax1.imshow(array1)
fig, ax2 = plt.subplots(figsize=(5, 10))
ax2.imshow(array2)

Notebook arguments

• --position 73% will set the slider start position to 73%.
• The height of the widget.
• --height 220 will set the height to 220 pixel.
• When --heightis not provided, the default height of the widget is 300 pixel.
• --height auto will set the height by the value of the first image's resolution in vertical direction.
• The widget's width will always be adjusted automatically.

Notebook formatting

Formatting with black can be done this way:

1. pip install 'black[jupyter]'
2. black --python-cell-magics splity splitview_magic.ipynb

Developer Installation

1. git clone --recurse https://github.com/kolibril13/jupyter-splitview (Note: In case that the repo was already cloned e.g. with the GitHub Desktop client, the GitHub submodule has to be loaded via git submodule update --init --recursive)
2. poetry install

Changelog

Milestones / Wishlist

• Handle cases where n ≠ 2 images. Currently: All further img are ignored.

• implement tests, find out how to test a magic class

• Idea: Second option without using cell magic:

from splitview import Splity # (does not yet exist)
my_splity = Splity(left_layer=img1, right_layer=img2)
display(my_splity)

• Make this work also in VSCode notebooks, see this issue.

• Some other nice views, like these:

Round Mask:

Double Round Mask, Second one with 50% opacity:

Gaussian Mask (no priority):

0.1.1

• Drop the github.com/NUKnightLab/juxtapose backend and replace it with github.com/Octoframes/compare_view.

• Implement Round Mask

0.1.0

• Update dependencies
• Update JupyterLite version
• Fix: in JupyterLite, a figure has to be explicitly called by plt.show()
• Better installation workflow

0.0.8

• Fixing problem with cell id and notebook reloading
• Experimentally lowering the dependencies to ipython = ">=6.0.0" and ipykernel = ">=5.0.0" so that jupyterlite will work hopefully.

0.0.7

• Rewrite of the import of JavaScript and CSS to make it more robust when closing and opening the notebook
• First attempt to add a JupyterLite example.

0.0.6

Fix poetry workflow

0.0.5

• Ship the javascript directly with the package, so no internet connection is required
• use jinja2 to save HTML in separate file
• load stylesheet and javascript only once in the beginning, and not in every cell that contains the splitview widget.

0.0.4

• New --height parameter

0.0.3

• default slider position
• updated minimal example
• internal code restructuring and formatting
• Handle import in non jupyter context

0.0.2

• save images in base64 strings and don't load images to disk (increases package security).

0.0.1

• First release