Accelerated large-scale image procesing in Python

A hands-on session presented by Joan Rue Queralt, with the collaboration of:

Matthieu Simeoni, Sepand Kashani, Thomas Debarre, Daniele Hamm and Salim Najib.

Content

• Presentation: Slides.
• Notebook 0: Data preparation. Downloads Hubble space telescope data and chunks it into small images.
• Notebook 1: Introduction to Numba and JIT compilation.
• Notebook 2: Introduction to Dask and the dashboard.
• Notebook 3: Introduciton to Dask-image and example of large-scale image processing.
• Notebook 4: Application of Dask + Numba: the structure tensor for feature extraction.

Note

Before starting, please clone this repository and install depenencies as follows:

$ git clone https://github.com/joanrue/accel-large-image-proc-talk
$ cd accel-large-image-proc-talk/
$ conda create -n accel_env python=3.11
$ conda activate accel_env
$ pip install jupyter
$ pip install graphviz
$ conda install matplotlib scipy numba scikit-image dask distributed dask-image nodejs zarr -c conda-forge
$ pip install dask-labextension
$ jupyter-lab