Drosophila automation

01 / Context

The problem

Preparing and observing fruit flies combines repetitive laboratory work with a difficult perception problem: very small, moving subjects whose appearance and position change. Useful automation has to connect what a camera detects with what the physical handling system can actually do.

02 / Ownership

My contribution

I implemented channel detection and climbing-assay vision, and co-developed classification, database, and interface components. I worked within a seven-person team; the specimen-handling mechanism, electronics, motion control, and broader integration were shared across teammates.

03 / Architecture

How it works

1. 01

   Locate

   Channel vision identifies subjects within the physical research workflow.

2. 02

   Coordinate

   A three-camera team system connects perception with motion, specimen handling, and Raspberry Pi control.

3. 03

   Observe

   Climbing-assay views and tracking summaries turn movement into information researchers can inspect.

4. 04

   Review

   Classification, stored results, and the interface support review of the experiment.

04 / Validation

Evidence & scope

• Implemented vision and assay subsystems within the team prototype.
• The public project site explains the architecture, contribution boundaries, and evaluation context.
• The team received the People's Choice Award.

05 / Perspective

Engineering takeaway

Image classification, detection, physical pickup, and a complete assay have different failure modes. Evaluating each stage separately makes an integration problem easier to understand than reducing the whole platform to one accuracy number.