Skip to content
D-CSIL

Project File

D-CSIL CANIS

A design for computational interpretation of canine communication from audio, video, pose, and interaction context.

ConceptA cool idea — still on paper, and labeled that way.

The Idea

THEORETICAL

CANIS asks whether multimodal machine learning could infer communicative intent across human-canine interaction — fusing vocalization audio, video, pose estimation, environmental context, and per-dog interaction history. Framed narrowly and falsifiably: do a dog's signals cluster into stable classes that predict subsequent behavior better than chance?

Honest Status

UNVERIFIED

This is a concept. No data has been collected, no models trained, no experiments run. It appears here because the lab publishes its questions as well as its results — and because the recording rig design borrows directly from hardware the lab already runs (Pi-class edge capture, Whisper-style audio pipelines).

Current Limitations

  • −No dataset, no baseline, no experiment — nothing beyond design notes exists yet.

Next Steps

  • +Define the pilot capture protocol and labeling scheme.
  • +Pre-register the first falsifiable within-dog classification test before collecting data.