Deploy ML Demos to Hugging Face Spaces with CI/CD
How to deploy a machine learning demo to Hugging Face Spaces in a fully automated way using a GitHub Actions CI / CD pipeline, so you never have to deploy by hand again.
by Martin Lellep
How to deploy a machine learning demo to Hugging Face Spaces in a fully automated way using a GitHub Actions CI / CD pipeline, so you never have to deploy by hand again.
A visual explanation of the WordDetectorNet model, which combines a neural network with DBSCAN clustering to find handwritten words on a page.
I built an AI app to help my mom transcribe audio embedded in PowerPoint slides.
A small red teaming experiment reveals how a single word can bypass content filters in Apple’s AI image generator.
Edinburgh buses are great but slow. I analyzed the data and found the issue - bus stops are too frequent, averaging just 323 meters apart!
I dedicated over 1,600 days to a data science project, only to see it fail. In this article, I share a new checklist I developed to help prevent similar setbacks in the future.
Slides and materials for my MLOps Scotland 2024 talk about Xournal++ HTR .
I made a mobile app using no-code to capture potholes in Edinburgh to help future cyclists.
I implemented an Online Handwritten Text Recognition ( HTR ) system using PyTorch, based on Google paper.
I share all of my learnings about going on a workation as a PhD Student to enhance productivity and creativity, including a cost breakdown.
Design of a custom case for your digital stylus. I use it every day and created it with FreeCAD.
Slides and materials for my mini Machine Learning course at Princeton University s Datta Lab.
My presentation notes for a seminar that I recently gave at the University of Edinburgh on dense representation learning for financial applications.
I used an app to capture my daily streching for 6 months and present the quantitative results in form of a figure.
Supplementary materials for my rC3 2021 contribution Optimising public transport: A data-driven bike-sharing study in Marburg .
Interactive versions of the Nextbike transition matrix figure from previous blog articles.
A quantitative update of the parking demand in Marburg during the Corona pandemic 2020.
In this article, I use machine learning to predict the number of parked bikes in two ways.
I use quantitative analyses to derive social and environmental benefits of the Nextbike system in Marburg. Also, I investigate which routes in Marburg are popular among Nextbike users. Hence, I offer quantitative arguments as to why bikes are good for Marburg and how to improve the biking experience in Marburg even further.
In this article, I draw quantitative conclusions for cyclists who use Nextbikes in Marburg. For these quantitative conclusions, I use Nextbike data that I previously scraped.
I collected Nextbike data in Marburg and introduce my plan to evaluate the data in this article. Also, I present what the data is made up of and how it is obtained in detail. Lastly, a first temporal analysis of the bike usage in Marburg is presented.
I show my small K40-based hobby laser cutter setup. I include one bonus tip: How to make money with a laser cutter.
A list of small designs like 3D prints, laser cuts and other constructions. Most of them serve a functional purpose. This list is continuously updated.
A laser cut community gift for Xournal++ developers.
I show how to use 3D scanning for CAD modeling and how I built a custom PC mouse box with the method that I present here. Neither do you need a GPU nor a powerful PC to follow along this tutorial. Also, the 3D scanning software is for free.
The parking demand in Marburg is analysed quantiatively based on publicly available data. In addition, Gaussian Processes are used for spatial predictions.
I built a wooden lamp around the Sonoff Basic IoT module.
My experiences with using my phone in Russian for 1 year.
Over are the days of juggling a bike lock while riding my bike!
A description of how to export Anki statistics by directly hacking the Anki desktop client.
Results of the experiment to learn Italian autodidactically for one month.
3D print and resin cast of a turbulent flow from 3D simulation data.
Results of 120 push ups for 30 days.
Application of greasing the groove training method to push ups.