bionictoucan / slic Goto Github PK

(he/him) Solar physics PhD turned medicines manufacturing postdoc with an interest in how machine learning can accelerate and benefit data processing pipelines. I am also interested in science and data science education and enjoy producing teaching materials.

N.B. Any deep learning mentioned below unless explicitly statred otherwise is construced using the PyTorch framework.

Medicines manufacturing repositories:

• SEssile Drop Video ANalysis (SeDVAn) 💧📹: an end-to-end framework for the analysis of videos from sessile drop experiments. This combines image segmentation, sequence-to-sequence learning and mechanistic modellling to characterise the absorption and swelling processes which take place during tablet disintegration.

Solar physics repositories:

• The Seeing AUtoeNcoder (Shaun) 🌤️ 👀 : A method for correcting the effects of the Earth's atmosphere on narrowband optical solar flare observations. This uses a fully-convolutional autoencoder to learn atmospheric seeing corrections based on a model derived from the statistics of turbulent media applied to data with minimal atmospheric distortions.
• crispy 🌞 🔭 : A Python package for working with imaging spectropolarimetric solar data in fits or zarr format. Designed originally for the Swedish Solar Telescope's CRisp Imaging SpectroPolarimeter (SST/CRISP) instrument, this package will work with any imaging spectropolarimetric data of the Sun.
• HYPerspectral Image Augmentation (Hypia) 🖼️: A Python package to apply data augmentation to hyperspectral images when training deep neural networks. This builds upon torchvision's transforms but makes it so that the channels dimension does not have to be 3.
• SoLar Image Classification using convolutional neural networks (Slic) 🌅 🤖 : A deep CNN trained to classify Hα images from Hinode's Solar Optical Telescope (SOT).
• RADYNVERSION 📈 : An application of an invertible neural network (INN) trained on simulations of solar flares to estimate the parameters of the flaring atmosphere from a set of observations.

Teaching Materials:

• Teaching: This repository contains a tutorial I gave to fellow PhD students about how unsupervised machine learning works and how to apply it in Python as well as an introduction to machine learning tutorial I presented at the Machine Learning in Heliophysics conference in 2019.
• Glasgow Machine Learning Course 2019: A course I co-created with a fellow PhD student to teach PhD students and postdocs how machine learning works, about the different kinds of machine learning and how it may be applicable in their research and how to go about implementing it in their research.