Vision & hardware · Public team project
AutoFrame
A camera that can reconsider its next shot.
A team vision-language photography prototype that observes a scene, retrieves previous attempts, proposes camera adjustments and closes the loop through a physical rig.
The question
Choosing a camera pose requires connecting visual interpretation to hardware action, then checking whether the next image actually satisfies the objective.
How it works
- Connect an Android camera server and a three-axis ESP8266 rig to the Python control loop.
- Use scene understanding and retrieved previous attempts to propose camera settings.
- Validate proposals, record the resulting image and apply stopping conditions across the iterative session.
Evidence & results
- The public team repository contains the control-system implementation.
- The established contribution record includes 167 prior attempts across 29 objectives and 28 autonomous sessions; these counts are user-confirmed.
- An existing public demo post shows the project.
Scope & next steps
- Team contribution; no claim of sole ownership of every hardware or software component.
- A previously quoted model-rated quality score is omitted because its scoring protocol is not sufficiently verified.
- README reference art is illustrative and AI-generated, not photographic evidence of the rig.
Go one level deeper.
Ask about the implementation, tradeoffs, or my contribution.
Explore with the portfolio guide