Picapike

Purpose and goal:

Sport fishing for pike is today conducted for the most part as catch and release. In social media, you see people posing with their catches before the fish are released back to their right environment.

Our objective was to create a fishing app where fishermen collect their catches and share the experience with each other. Our app identifies pike at the individual level using AI. When we aggregate catch data, we can contribute to research and management. All catches and recaptures that are registered give us a greater understanding of our ecosystems and aquatic environments.

Expected results and effects: As this year´s sport fishing fair has been canceled due to Covid-19, we are making a soft launch via our website.

We will start generating data now that the ice is clearing and fishermen can start fishing for pike. When live data begins to flow into the system, we will be able to refine our AI model. The system can already recognize fish and suggest matches. However, the result depends on a number of factors. More images will refine the system´s capabilities. At the same time, we will be able to begin to aggregate data for the best interest of both researchers, fishermen and the fish populations.

Approach and implementation:

At the start of the project, our AI model was validated for a relatively tight set of parameters. We discovered during the summer / autumn that the model could not deliver adequate answers in the more open question that is the case when the system is in operation. As a result, we had to develop and train a new AI model based on these experiences.

In addition, we had problems with the AI platform from Scaleout Systems. It became clear to us that this was developed for distributed systems and did not fit well with our AI model in the cloud. Hence the decision was made to develop our own platform.