Teaching a small drone to see
A vision model on a tiny board, built over one weekend with my group's help.
Running vision in the cloud added half a second of lag, and at that speed a drone has already hit the wall.
What we moved to the board
A small detection model, cut down to run at twenty frames a second.
A simple planner that only trusts the last three frames.
A safe landing that needs no network at all.
Edge AI forces you to decide what really matters. That was the best lesson of the month.
- Robotics
- Edge AI
Related writing
BlogI built my first voice agent in a weekend
What broke, what surprised me, and why latency matters more than the model.
BlogFive prompts that fixed my RAG answers
Small changes that stopped my assistant from guessing, with no new tools.