Skip to content

For Drosophila research labs

Automated scoring for fly phenotype assays

Upload footage from a climbing assay or a phenotype photograph, and get counted, scored results ready for review and export.

See it in action ↓

Prototyped around assays used in the neurodegeneration and cancer research programs at the Drosophila Research & Training Center, Ibadan.

Worked example

See how a batch gets scored

A climbing-assay batch, detected and scored end to end. The data shown is illustrative — running it on your own footage is part of the pilot.

A climbing-assay vial with six flies detected and boxed. Five carry confidence scores between 0.88 and 0.98; one is flagged at 0.71 for manual review.

How scoring works

Two systems, two different jobs

01

Detection

Frame-differencing and blob tracking locate and follow each fly across the video. This stage is deterministic: the same footage produces the same count on every run.

02

Scoring

A vision model evaluates phenotype severity — eye roughness, disc size, wing shape — against reference grades calibrated to the assay.

03

Review

Low-confidence results are flagged with a reason, such as occlusion or an ambiguous frame, and held for a quick manual check before export.

Built for two of DRTC’s core assays

Coverage for the phenotypes most published fly work already depends on.

Climbing assays

Negative-geotaxis tracking from a handheld phone video, no rig required.

Eye phenotype scoring

Roughness and ommatidial disorganization, graded against reference severity.

Wing and disc scoring

Size and shape measurements taken directly from dissection photographs.

Export formats

CSV output alongside a caption drafted in standard methods-section phrasing.

Built for two moments

Who this is for

Score a full batch in roughly the time it takes to count three flies by hand. Low-confidence results are flagged before they reach your spreadsheet.

  • Phone-video input — no equipment purchase required
  • A confidence score on every individual result
  • Flagged rows held for manual review before export
~30×faster than manual scoring
0 sec

Median time to score one batch

0

Flies scored per batch

0%

Average detection confidence

0

Flagged for manual review

Figures reflect the illustrative example on this page, not measured results from a live pilot.

Join the waitlist for early access

We’ll reach out when the pilot is ready to run on your own footage.

Prefer email? Reach out directly at ayanabrand@gmail.com