The other day remembered the ancient project, where we had to blink out the camera's firmware with its own led (https://cutt.ly/urgKrmi). As the firmware was dumped and analyzed, people began to write modding software for those cameras, which eventually grew up not only to the CHDK project (https://cutt.ly/YrgKphE) but also to a magiclantern.fm project. From that story, I still like a couple of cool tricks, like measuring the shutter speed using a Dremel drill (https://cutt.ly/zrgKj9m) and memory dump transfers via variable power consumption (https://cutt.ly/YrgKlBd). As for me, I ran some demoscene scroller on my camera (https://youtu.be/bOyJTHC1cBc) and was distracted by something else.
Also: a non-official opensource firmware project for soldering irons TS100 (https://github.com/Ralim/ts100) and the firmware modding project for Convoy S2+ flashlights (http://cholla.mmto.org/lights/convoy/mods/). It's funny, that it can blink out its temperature in degrees Celsius (https://bit.ly/35F37Dr).
- RL-sandbox for training algorithms on the assembly of IKEA furniture (https://github.com/clvrai/furniture).
- Phonetic analysis and description of typical sounds teenagers make (https://www.youtube.com/watch?v=ZY2R_K3NFPo).
- A webcam in a favicon (https://twitter.com/davywtf/status/1119783380734836737) (reminded me a bit of my old project (https://www.youtube.com/watch?v=6LO9fFAGtf0)).
- A museum of funny and silly names of wi-fi networks (http://museumofwifi.com/) from around the world. Also: museums of soda machines (http://vgsmproject.com/), manholes (https://videogamemanholes.tumblr.com/) and statues of Moai (https://moai.games/) in computer games.
- A neural network for generating a knitting pattern from a photograph (http://deepknitting.csail.mit.edu/) (reminds me my ancient post on the math knitting (https://medium.com/altsoph/math-knitting-97308ca3595e)).
- Averaging the flags of different continents (https://www.reddit.com/r/dataisbeautiful/comments/85s2nl/average_flags_of_the_world_2_mean_and_mode_flags/) and drafts of the flags of the planets of the solar system (https://www.reddit.com/r/vexillology/comments/cbnkmw/flags_of_the_eight_planets_plutocharon_and_planet/).
- Misha Grankin shared the Russian pre-trained GPT-2 model (https://github.com/mgrankin/ru_transformers/#3-id-like-to-download-your-models).
- GPT-2 generates step-by-step instructions for drawing, a special script turns them into pictures in the style of Levitt (https://twitter.com/DrBeef_/status/1177012430230183937).
- GPT-2 generates svg icons and fonts (https://www.fluate.net/en/travaux/vectoglyph).
Peaceful ML on duty for humanity:
- A lip-reading by a neural network is a rich topic, I believe (https://github.com/Fengdalu/LipNet-PyTorch).
- A fast.ai student's coursework dedicated to automatically determining the degree of a building completion by space/aero photos of Zanzibar (https://forums.fast.ai/t/share-your-work-here/27676/577).
- An article "Detecting Kissing Scenes in a Database of Hollywood Films (https://arxiv.org/abs/1906.01843)".
- The history of the dispute between Symbolists and Connectionists in one understandable picture (https://twitter.com/IntuitMachine/status/1200796873495318528).
- A small but pretty dataset with national visual patterns (https://www.kaggle.com/olgabelitskaya/traditional-decor-patterns) -- Gzhel, Khokhloma, etc.
Interesting links from readers:
- Kirill Demura sent a link to the video of recording analog audio on three-inch floppy disks (https://www.youtube.com/watch?v=Xpr7B-7BFP4).
- Max Beketov shared an article about assembling circuits from logic gates based in Minesweeper (http://www.formauri.es/personal/pgimeno/compurec/Minesweeper.php).
- Pavel Lebedev gave a link to a kickstarter project of an electronic spinning top, which spins without stopping for hours (https://www.kickstarter.com/projects/fearlesstoys/limbo-the-magical-top-that-spins-for-hours).
- Bulat Karimov sent a link to a post (Russian only) where people consider the "Hitler index" for various articles on the Russian Wikipedia (https://pikabu.ru/story/kogda_nechego_delat_na_kanikulakh__igra_v_vikipedii_7129430).
Boreslav just sent me a great quote from Mark Twain's "Life on the Mississippi" as a follow-up on my recent post:
"...The Mississippi is remarkable in <...> its disposition to make prodigious jumps by cutting through narrow necks of land, and thus straightening and shortening itself. <...>The town of Delta used to be three miles below Vicksburg: a recent cutoff has radically changed the position, and Delta is now _two miles above_ Vicksburg.
...A cut-off plays havoc with boundary lines and jurisdictions: for instance, a man is living in the State of Mississippi to-day, a cut-off occurs to-night, and tomorrow the man finds himself and his land over on the other side of the river, within the boundaries and subject to the laws of the State of Louisiana! Such a thing, happening in the upper river in the old times, could have transferred a slave from Missouri to Illinois and made a free man of him..." (http://www.gutenberg.org/files/245/245-0.txt)
The meander movement of the Ucayali river in Peru over 30 years. Found on Twitter https://twitter.com/geogabout/status/931996572434526208 #meander #river #timelapse
And since I mentioned NaNoGenMo, here is my personal list of somehow funny or interesting entries:
* Bobey Dig (https://github.com/NaNoGenMo/2019/issues/96) -- reading Moby dick with a stuffy nose (by adjusting the phonetic library Pincelate to remove nasal features, turning down fricatives, etc.).
* Personalized Bibles (https://github.com/NaNoGenMo/2019/issues/129) -- replacing the names of the characters to names of pokemons or names from books about Harry Potter.
- A Most Portentous Dream Diary (https://github.com/NaNoGenMo/2019/issues/148) -- the author compiled a grammar for Inform 7 generates a dream diary.
- Knowledge is power (https://github.com/NaNoGenMo/2019/issues/113) - a text from the first sentences of different Wikipedia articles, compiled so that each following sentence "explains" the previous (last word).
- The Orange Erotic Bible (https://github.com/NaNoGenMo/2019/issues/18) - GPT-2 network first read erotica a lot then it was asked to continue some pieces of the Bible.
- I myself participated this year with the Paranoid Transformer model (https://github.com/NaNoGenMo/2019/issues/142).
It's kinda boring to write about the same things again and again, but I don't want you to miss some things, so here is a general list of non-news I somehow mentioned before in my telegram channel (https://t.me/pathetic_low_freq):
* The annual Christmas lecture from Donald Knuth recently published (https://www.youtube.com/watch?v=s_kd7xE-HqA).
* The AI/ML Creativity Workshop (https://neurips2019creativity.github.io/) is taking place on NeurIPS today (the same one, on which we presented our album Neurona two years ago (https://medium.com/altsoph/neurona-ep-da4f48a23295)).
* Books gradually appear on Amazon in official co-authorship with GPT-2 (https://www.amazon.com/s?rh=p_27%3AOpenAI+GPT+2).
* Video of talks from Strange Loop 2019 (https://www.youtube.com/channel/UC_QIfHvN9auy2CoOdSfMWDw/videos) and Roguelike Celebration 2019 (https://www.youtube.com/playlist?list=PLi7jNGNQhwdg9M2K3s6W73E0BlWF0qrLY) are already published.
* The annual NaNoGenMo (https://nanogenmo.github.io/) took its place.
This year the annual text generation competition (https://nanogenmo.github.io/) had an interesting twist — Nick Montfort (@nickmofo) run a spin-off competition, Nano-NaNoGenMo (NNNGM), with only one additional rule - the program should have 256 bytes at most.
So I managed to craft the 123 bytes Perl script for a Markov Chain text generation (139 bytes in total with a shell command), check the details here https://medium.com/@altsoph/nano-markov-textgen-b80e1212f3b3 #NNNGM, #Perl, #obfuscation #text #generative
The loudness of typical songs by genre, visualized as raincloud plots, based on data from some Spotify's kaggle competition. The source is a reddit https://www.reddit.com/r/dataisbeautiful/comments/e4xk6q/music_loudness_by_genre_oc/. There is also a discussion about "the loudness war" and a huge database of 140K albums dynamic ranges (http://dr.loudness-war.info/).
Also check: samples of all sorts of instruments from http://www.philharmonia.co.uk/explore/sound_samples/ and a couple of selections of strange cosmic sounds from NASA https://soundcloud.com/nasa/sets/spookyspacesounds, https://soundcloud.com/nasa/sets/solar-system-beyond-sounds.
#sound #datasets #dataviz
An old (2013) Japanese rapier combat visualization project called “Fencing Visualized” (http://dentsulab.tokyo/works/fencingvisualized). It's not in realtime, and moreover, it seems that special suits with sensors for motion capture were used. But still, it's beautiful, and soon, thanks to all sorts of PoseNet (https://www.tensorflow.org/lite/models/pose_estimation/overview), DensePose (http://densepose.org/) and BodyPix (https://blog.tensorflow.org/2019/11/updated-bodypix-2.html), it will be possible to do such things simply based on the video stream. #visualization #fencing #cv
A histogram of frequencies of world metro/underground lines colors (hue) by Xan Gregg.
* the discussion thread on Reddit https://www.reddit.com/r/dataisbeautiful/comments/7smt2w/all_the_worlds_metro_line_colors_by_hue_angle_oc/,
* the data https://github.com/xangregg/data/blob/master/world_rail_line_colors.csv,
* a previous similar project http://web.archive.org/web/20180323190620/https://www.c82.net/work/?id=355.
Also check 200+ logos of world's subway systems, sorted by similarity: http://mic-ro.com/metro/metrologos.html?size=lg&sort=similarity.
The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!