challenges and opportunities for ML, citizen science and fish research

[AI for fish]

Here are some of the challenges and opportunities that emerged from the Day 1 of the workshop. We have summarised them here in five broad categories. Please feel free to respond and share your ideas on how we can overcome the challenges and make the best out of these opportunities: 

1.Effective angler and citizen engagement 

• many local apps

• can we have one global app like ornithologists do? 

• can we link some apps through data and tool sharing?

• can we learn from other programs what motivates people to engage? 

(tournaments, specific questions with clear goals, gamification)

2.Expert identified fish photos for algorithm training

• can we share expertise across groups? 

• outsource to the scientific community?

• how should contributors benefit? 

• what tools do we use?

(Zooniverse, FishBase experts?)

3.Which algorithms and methods to use? 

• efficient open source pipelines for annotation, image augmentation and ML training

• sharing publications, protocols, methods

• shared publication & report about the most efficient publicly available tools 

4.How do we use the data? What does it say? 

• what can we communicate back to anglers, citizen scientists and managers?

• how can it help to maintain healthy fish stocks? 

5.How to ensure the data is used widely and doesn’t get lost? 

• best approaches for data sharing while accounting for IP, privacy, commercial interests? 

• how can this data sharing benefit each app/organisation?