Any Tensorflow Lite experts out there? I'm trying to do object detection, but object_detector.DataLoader.from_csv only loads about 600 of my 2000 bounding boxes 😭
I thought it might be a malformed CSV, so I tried shuffling the rows, but I always get a training dataset size exactly equal to my number of images, and test + val datasets about half that (my actual split is 80/10/10)
I've also tried pushing max_num_images up explicitly, but it doesn't seem to have any effect
oh by the way, the above toot inspired by the amazing talk by @jimjamslam I just caught up on. His wonderful slides here: https://runapp2022.talks.jamesgoldie.dev/
My lovely partner got me a sack that has the #rstats graphics device shapes printed on it and of course I'm in love with it
Very thankful to the lovely folks at Runapp for letting me do a deep-dive on #Quarto and how it's changing the way I do data analysis and #datavis!
I'm really excited about this tool and am keen to see what it does as it moves toward a 1.0 release in the next few months.
➡️ Slides: https://runapp2022.talks.jamesgoldie.dev
I’m talking #quarto next week! If you write reports, blog posts, notebooks or slides in #rstats, #python, #julialang or #javascript, grab some lunch and tune into this talk!
R users, learn how Quarto brings tools together from across the RMarkdown ecosystem, along with some new tricks!
Python and Julia users: learn how you can get in on this "reproducible reporting" thing you keep hearing about, using existing Jupyter notebooks or using plain text docs that work well with version control! (You don't even need R 😉)
Everyone: learn how throwing Observable into the mix can improve your docs and slides even more, letting you add menus, sliders, web visualisations and more!
If you've ever worked in a #datascience language like R or Python and thought, "Yeah, but I really don't wanna touch JavaScript", I reckon Quarto and OJS are the easiest, most tempting way we've ever seen to try it out.
It’s a free talk! RSVP now: https://www.eventbrite.com.au/e/quarto-create-beautiful-documents-with-r-python-julia-and-observable-tickets-333340770217
This article from Jeff Masters talks about resurgent #heatwave risk in India this month: https://yaleclimateconnections.org/2022/05/india-and-pakistans-brutal-heat-wave-poised-to-resurge
It highlights a paper I missed in Jan and testing Sherwood & Huber’s proposed theoretical 35°C wet bulb limit a decade ago. It's a small study, but it finds young, healthy people failing to adapt at wet bulbs around 31°C—and even lower, in super high dry temps (eg. around a 26°C wet bulb in 50°C, lower humidity conditions).
Would love to see this work repeated with more people and eventually extended to investigate heatwave effects! 👏🏻
📄https://journals.physiology.org/doi/abs/10.1152/japplphysiol.00738.2021
Where're my #rstats folks on Mastodon? 🥳
If you don't know me, I do #climatechange communication, a bunch of #datascience-y and #datavis stuff, and I used to be a climate and health researcher 🥳
It is proven! I am rensa on Keybase: https://keybase.io/rensa/sigchain#abdd0986903bd5496d782f617a6e9a113701d95e8e9264021e6dd00f81f220a10f
Data and Digital Storytelling Lead @ 360info.org (personal views here). Former #climatehealth; making things for fun and social good ♥ He/him