Initiative #165

Description de l'initiative Non traduit · origine : anglais

The AIMMS project develops a platform to automate and accelerate the interpretation of aerial imagery for marine mammal population assessments. By streamlining annotation and review, and enabling scientists to develop custom machine-learning models without advanced coding skills, the AIMMS workflow significantly reduce processing time, accelerate delivery of science advice, and lower operational costs.

Description du problème Non traduit · origine : anglais

DFO Science must provide timely science advice on the population size of harvested and at-risk species using aerial imagery. Increasing volumes of high-resolution imagery now exceed current human resources and internal capacity for manual annotation. As a result, data processing is delayed, science advice for priority species is slowed, and financial costs are rising. DFO currently lacks a scientific tool that efficiently analyzes survey imagery and provides estimates of marine mammal abundance.

Résultat escompté si l'IA est retenue Non traduit · origine : anglais

Success means implementing an automated workflow for image annotation and analysis using an open-source, user-friendly machine-learning platform, freely available across DFO. Fully operational models, where predictions are accepted as final, could reduce aerial imagery analysis time by an estimated 72% compared to manual annotation. This would allow experts to focus on higher-value work, clear processing backlog, and accelerate delivery of science advice for priority species.

Risques Non traduit · origine : anglais

Delays in software development and integration may arise if technical challenges occur while enhancing Label Studio’s features and usability. Furthermore, limited personnel with the necessary expertise or unexpected staff turnover could affect critical workflows, including annotation and model development. Adoption also poses a risk; if key DFO science teams are slow to engage or face challenges with training, the impact and deployment of the enhanced platform may be reduced. To mitigate issues related to software development and integration, the team will collaborate with experts from CDOS and/or OCDS, with regular check-ins to track progress and address issues promptly. Frequent testing and an open feedback loop with developers and users will ensure challenges are quickly identified and resolved, keeping the project on schedule. To reduce the risk of not having enough skilled staff or people leaving the project, we will make sure team members are trained in multiple areas and maintain easy-to-follow instructions and guides. Moreover, we already have an established project team that is ready to continue the work with the expertise that they have gained over the course of the project.

Alignement stratégique

• Amélioration des avis scientifiques pour les pêches

Domaines d'impact

• Temps

Faisabilité

• ✓Les données sont disponibles et utilisables