Computer-vision research and development for the UBC Uncrewed Aircraft Systems team's 2027 Aerial Evolution Association of Canada student competition entry.
The current focus is Task 1, the Herd Survey: detect and count deer decoys, group them into spatial clusters, read ear-tag IDs, identify anomalous interactions with human artifacts, and support timely generation of the competition report.
AGENTS.md: durable repository instructions for teammates and AI tools.CONTRIBUTING.md: lightweight contribution and handoff workflow.docs/README.md: documentation map explaining what each file owns and when to update it.docs/competition/task-1.md: confirmed Task 1 requirements and the current CV decomposition.docs/competition/ambiguities.md: unresolved competition and engineering questions.docs/experiments/README.md: how to preserve useful experimental evidence.docs/datasets/README.md: dataset documentation expectations.
dataloaders/: reusable loaders for core datasets so training and evaluation use consistent behavior.prototyping/: individual model prototypes and early investigations.docs/competition/: competition sources, interpretations, and open questions.docs/experiments/: durable summaries of meaningful successful, negative, and inconclusive experiments.docs/datasets/: dataset provenance, licensing, versions, splits, and limitations.docs/decisions/: consequential technical decisions and their rationale.
Large datasets, flight recordings, trained weights, and generated outputs should not be committed directly unless the team adopts an appropriate storage policy.