COASTVISION - Computer vision to expand monitoring and accelerate assessment of coastal fish

CoastVision seeks to develop computer vision for efficient processing of video from coastal surveys.
This approach will be applied to case studies for detection, species classification, sizing, tracking
and re-identification (re-ID) of individuals for enhancing monitoring and stock assessment of coastal
fish populations.

Secondary objectives:
1) Improve our current models for detection and classification of coastal fish such that they achieve state-of-the-art accuracy, trained and verified on data from several surveys.

2) Expand the automated pipeline to include sizing.
3) Apply a new camera system for automated collection of data for re-RD using radio frequency identification.

4) Train and validate re-ID models with a target of > 90 % accuracy over short periods (<6 months) for cod, corkwing and ballan wrasse.

5) Apply the CoastVision pipeline to on-going surveys of fish populations and obtain key data for
stock assessments, such as growth and fishery selectivity functions.