---
title: Project Ideas
layout: default
permalink: /project_ideas/
excerpt: |
  A semi-hidden page to jot down ideas for projects.
mathjax: false
---

# Project Ideas

* Write some new firmware for the [DMM SAO](https://github.com/flummer/dmm-sao/blob/main/firmware/code.py) I got at the hackaday supercon.

## 3D models of vertebrates
https://www.morphosource.org/
https://sketchfab.com/FloridaMuseum/models
https://www.morphosource.org/concern/media/000727033?locale=en
https://www.morphosource.org/concern/media/000658134?locale=en
[Triangulation Theodolite By Hildebrand Freiberg, 1902](https://www.morphosource.org/concern/media/000657801?locale=en)
[Cool Green Theodolite](https://www.morphosource.org/concern/media/000655681?locale=en)
[Nice Lumbar](https://www.morphosource.org/concern/media/000532571?locale=en)
[Shark Head](https://www.morphosource.org/concern/media/000677539?locale=en)

## Ferrofluid music thing

Put that ferro fluid in a round bottom flask and put a coil nearby

## Battle ships

What's the optimal battle ships placement and guessing strategy?
Presumably the best battle ship placement strategy would be where you cheat and break your ships up into 1 pixel ships and then place them randomly.

I guess the second best should be where you place them randomly using rejection sampling (try to place in random position, if you intersect with a previous placement try again).

What's the best guessing strategy? If you're up against the cheater presumably you can't do better than random. But if you're up against the random but physically correct player, I guess the shape of the ships will rule out some configurations and give you and advantage.

If you're up against a human placer then presumably you could do better by exploiting human placement patterns.

Idea: Train a pixel-pixel model to take in a three state battle ship board where the cells are (Hit, Miss, Unknown). Draw examples from the physically correct and randomly mask them. Train with (x = masked board, y = unmasked boards) to try to get a prediction model for likely placement of ships given known squares.

Extension: If you have a good prediction model, do you just choose the square predicted to be mostly likely to have a ship? Or do you iterate over every square and somehow measure how much knowing the value of that square would improve our confidence?

## Laser Cut Projects

- Laser cut things from [here](https://blogs.loc.gov/maps/2025/03/globe-gores/)

## Thermal Camera Projects

- Pair camera with iPad and fuse the 3D model generated by the iPad with thermal input.
- Use Raspberry Pi + screen + camera + thermal camera + (depth sensor?) to fuse thermals and camera input.

> - One approach requires learning iOS app development but may be more useful in the long run.
> - The other needs more hardware but could be fun for other projects too.

### Additional Ideas

- Use DeepL to segment the image input, overlay the thermal image, then do a best-fit to impute segment temperatures.
- Track motion in the camera input and overlay thermal input (copy what the iPad does). Potentially use OpenCV.
- Can the Raspberry Pi GPU be used with compute shaders yet?

## Getting Into SMD Manufacturing

- Start with a simple test board — maybe an ESP32 datalogger.
- Get tools.
- Decide how to heat: maybe a DIY hotplate?

### DIY Hotplate

- Already have the hotplate.
- Do I have the switches though?

## Custom Home Assistant / Smart Speaker

- Use existing Raspberry Pi.
- Add Home Assistant.
- Add air quality sensor.
- Add RGB LEDs for display.
- Add a small screen for extra metrics.
- Make a nice white case with plenty of space.
- Maybe add a microphone and try to find an open source smart speaker:
  - [Mycroft AI Docs](https://mycroft-ai.gitbook.io/docs/using-mycroft-ai/get-mycroft/linux)
  - [ODAS on GitHub](https://github.com/introlab/odas)

## Other Ideas

- Replace driver board on Ender 3 for quieter operation.
- Hook up PSU control with a transistor.

## Build a Linux System from Scratch

- [Linux From Scratch Guide](https://linuxfromscratch.org/lfs/view/stable/)

## Adjustable Bedside Lamp

- Adjustable colour temperature and brightness.
- Use high CRI LEDs and RGB LEDs.
- Power with 20V USB-C supply.
- Add microcontroller and rotary encoder.

## Custom Low Profile Keyboard

- [Hackaday: Low Profile Keyboard Guide](https://hackaday.com/2022/03/16/a-clear-guide-for-a-low-profile-bespoke-keyboard/)