I trained a convolutional neural network with no dependencies @sphaerophoria
I trained a convolutional neural network with no dependencies  @sphaerophoria
Uploaded August 2025 | Updated November 2026, 18 hours ago
My food tracker needs a barcode scanner, my barcode scanner needs a neural network, my neural network needs to be trained, training needs convolutions, convolutions need gradients, ow ow ow ow.

Please consider supporting. This content WILL end some day, but every dollar I make pushes that day further out

Join on youtube for happy hour vods youtube.com/channel/UCXzL31BCxf8En1KT34gSK6g/join
Or on patreon patreon.com/sphaerophoria
Or on twitch twitch.tv/sphaerophoria
Or sponsor me on github (no benefits) github.com/sponsors/sphaerophoria

00:00 Intro
07:40 Training loop/Simple case
01:11:40 Try to learn barcode orientation
I trained a convolutional neural network with no dependenciesBut like, how well does it actually work?Arc raiders crafting/savlage/recycling utilitymalloc is for babies and losersWriting code to generate more codeWe always get distracted IMMEDIATELYZigs new IO interface is pretty neatPut the tiles in the box little broIm a zig contributor (again)!My color correction isnt working good...Wayland keyboard input handlingArbitrary GPU compute with OpenCL
sphaerophoria |

I trained a convolutional neural network with no dependencies

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER