Tim's Blog

I finally got around completing a long pending project idea: A diffusion image generation model implemented on a microcontroller.

Now, over the course of more than a decade, I spent quite some time manually investigating WS2812 addressable RGB LEDs and their clones.

Building wireframe polyhedra made from LED filaments, using graph theory to devise geometry and driving strategies.

I relocated my blog to Hugo due to easier maintainance and more control over content and layout.

Combining a deep-depthwise CNN architecture with variable quantization in BitNetMCU achieves state-of-the-art MNIST accuracy on a low-end 32-bit microcontroller with 4 kB RAM and 16 kB flash.

Todays candles have been optimized for millenia not to flicker.

LED-based festive decorations are a fascinating subject for exploration of ingenuity in low-cost electronics.

Going along with implementing a very size optimized neural network on a 3 cent microcontroller I created an interactive simulation of a similar network.

Bouyed by the surprisingly good performance of neural networks with quantization aware training on the CH32V003, I wondered how far this can be pushed.

The CH32V203 is a 32bit RISC-V microcontroller.