This project uses image recognition models to identify fish being cleaned at fish cleaning tables around NSW. The project is funded by the NSW Rec Fishing Trust and is a collaboration between Macquarie University, NSW Fisheries and a tech company called In2iti0n. We run several models both onboard the camera and in the cloud. The system first recognises if a fish is present in the video footage, then tracks each fish and takes a series of photos which are later uploaded to the cloud to be classified using a machine learning model. The data is collected and processed in real time and we have a variety of dashboards that can be tailored to suit data analysis / graphical output needs. We are focussing on the top 15 most popular rec fish species in NSW, but the learning algorithm can theoretically be applied to any location. For example, there is interest in this project from NT Fisheries and Tasmania. While the model can detect fish generally, accuracy improvements can be gained by training purpose-built model for specific locations. A combination of stereo vision techniques and IR dot projection also estimates fish length. An additional camera monitors boat deployment and retrieval at adjacent boat ramps providing information on infrastructure use. While our focus is mostly on gathering catch data for fisheries management, the model could also be handy for compliance. We are also trialling the tech at the Sydney fish markets to facilitate data gathering in that context which is currently both labour intensive and time consuming. While we are focussing on applications port-side, the system could also be adapted for boat deployment if needs be.