Hey Hunters! Today my colleagues and I are releasing Kangas, our open source library for exploring multimedia datasets, on PH 🚀 We're a team of researchers and engineers who work in MLOps, and over the last few months, we've been working on a tool that would make it easy for us to download a computer vision dataset and immediately begin exploring it. We didn't want to write a million utility methods to visualize an image or plot a chart, we didn't want to glue together a bunch of tools to query our data, and we didn't want to bump into memory limits every time we tried to load a larger dataset. We started working on Kangas as small project to scratch our own itch, and it quickly turned into a bit of an obsession. We released the library as part of its open beta two weeks ago, and we've been excited by the initial feedback. Kangas has racked up nearly 500 GitHub Stars, been explored by data scientists and engineers at a bunch of amazing companies (AWS, Tesla, and Meta, to name a few), and has already had some community contributions. We've included several example projects/quickstarts on the Kangas repository, and we are more than happy to answer any questions either here or on the repo as you get started exploring Kangas!

Kangas - Explore computer vision datasets in seconds

Replies