PhD Student at ML-Labs UCD | Using satellite images and deep learning to monitor the Irish coastline 🌊
Writer for Towards Data Science | I write about IML, XAI, Algorithm Fairness and Data Science in general ✍️🤓
Got to present some early work at #PIERS2023 in the beautiful city of Prague.
This work provides a link between the data and interpretations of segmentation models. We hope it will also provide insight into which indices can be used to automate annotations for future models.
Read about the dangers of gender recognition software, inadequate medical models, and the amplification of transphobic content. As workers in tech, you have the power to push back against these trends.
All medium partnership funds from this article will be donated to TGEU. Thank you to
@towardsdatascience for all the feedback.
No paywall link:
https://towardsdatascience.com/unmasking-ais-detrimental-effects-on-the-trans-community-d8f870949d79?sk=0b9c4036da9b16d884297bb93f12ed95
The rise of powerful models and accessible AI interfaces has made working with audio and music exponentially more streamlined, and new horizons continue to open up every day.
From music-tagging AI systems to advanced transcription workflows, discover new frontiers in audio machine learning with contributions by Leonie Monigatti, Ana Bildea, Max Hilsdorf, and Luís Roque. https://towardsdatascience.com/new-frontiers-in-audio-machine-learning-6474ffaa5cb9
I attempted to make a Timeseries gif using Landsat data. I used images of Howth and Bull Island in Dublin from 1980 to 2023. The first attempt was well 👇
Problem 1: I cropped the images but didn't realise positions would change depending on the image.
Problem 2: clouds!!!
Problem 3: what's with the black lines on some of the images?
As I learn things for my PhD, my content is starting to shift towards remote sensing. This article is the first of many :)
Learn how to deal with
- multiple spectral bands,
- large pixel values and
- skewed RGB channels
when visualising satellite images
No-paywall link:
https://towardsdatascience.com/visualising-the-rgb-channels-of-satellite-images-with-python-6d541af1f98d?sk=d43a1b06c6c2e9df0d1556e013ca58d5
SHAP is the most powerful Python package for understanding and debugging your machine-learning models. With a few lines of code, you can create eye-catching and insightful visualisations :)
We walk through the code used to calculate and display SHAP values. This includes explanations of:
- Waterfall plot
- Force plots
- Mean SHAP plot
- Beeswarm plot and
- Dependence plots
My brother has been doing coding interviews. I'm convinced these things have a material negative impact on the global economy...
For months, forcing highly skilled, highly paid workers to solve problems that have been solved millions of times before
Puts me off wanting to be a software developer 🤷♂️
This one is a deviation from the normal data science stuff. It is about the sources you can use when coming up with article ideas.
But, really, it is about my writing journey to date.
No-paywall link:
https://writingcooperative.com/never-run-out-of-article-topics-5-reliable-sources-fb161a77de3e?sk=0e9f629dd80ed0865189446f5636314e
I'm turning a few of my course videos into YouTube videos. This is the first of them.
I discuss how to interpret SHAP values and the package's applications. Let me know what you think :)
#SHAP #MachineLearning #DataScience #IML #XAI #Interpretability #Explainability
Autoencoders can reconstruct entire images. So, surely, they can be used for semantic segmentation tasks?
They can! But, we are able to get better results with U-Net. The key is the encoder and decoder no longer have to be separate. We can join them using skip connections. Together, these 3 components extract and localise features in image data.
Learn how this allows us to do accurate segmentation with less data (no-paywall link):
Towards Data Science
