Illuminating the global fishing fleet through machine learning

Опубликовано 30 сентября 2021, 16:00

• Blog post → goo.gle/2ZJQyth
• Notebook used to host the codelab → goo.gle/3iiT8g4
• Global Fishing Watch website (existing project) → goo.gle/2X1VJmT

Description:
Oceanic ecosystems are threatened by a lack of science-based fishery management, but machine learning can help. Welcome to People and Planet AI where we explore how machine learning helps us learn about the biological variety and variability on earth. In this episode, we show how Global Fishing Watch uses Google Cloud tools and machine learning to determine when a ship is or isn’t fishing. Watch to learn how you can create your own time series classification app that includes latitude and longitudinal data!

Chapters:
0:00 - Intro facts on unsustainable fishing
0:23 - Intro to Globalfishingwatch.org
1:07 - Series intro
1:30 - Meet Brian Sullivan Co-Founder of Global Fishing Watch at Google
3:11 - How Google is helping
4:13 - Architecture review for making a time series classification model
4:55 - What is Maritime Mobile Service Identity
5:14 - Converting irregular GPS signals to hourly timesteps
5:24 - Using a 1-Dimensional Fully Convolutional Network
5:40 - Fishing labels
6:03 - Training and evaluation datasets
6:20 - Defining the model
6:36 - Options to host model
6:42 - View of model making predictions in Colab notebook
7:01 - Try out the code sample
7:30 - Brian’s fun fact about the project

Global Fishing Watch:
• Global Fishing Watch on YouTube → goo.gle/3necA0Z
• Global Fishing Watch on Twitter → goo.gle/3nlk02m
• JOIN the Global Fishing Watch Team → goo.gle/3yXVOFp
• Labeled dataset on GitHub → goo.gle/2Vro0mn

Google resources:
• How to start a Google cloud project - $300 free trial link → goo.gle/3uawMlS
• Free course by Google on Data Preparation and Feature Engineering in ML → goo.gle/3tqYwCa
• Data normalization → goo.gle/38SMZCd
• Folium (open source Python library for interactive maps) → goo.gle/3kSHTNi
• Vertex AI → goo.gle/2XEGrEQ
• Deploying model as endpoint on Vertex AI → goo.gle/3kyEeUo
• Dataflow (data processing) → goo.gle/2XOYZTs
• Apache Beam → goo.gle/3B6pwd6
• Colab notebooks → goo.gle/3CTYLJy
• Cloud Run (web app hosting) → goo.gle/3CCyIXb
• TensorflowKeras → goo.gle/3CVC2gf
• Hosting a website on Google Cloud using Cloud Run → goo.gle/38QoIg8

Watch more
• People and Planet AI playlist→ goo.gle/PeopleAndPlanetAI
• Subscribe to Google Cloud Tech → goo.gle/GoogleCloudTech

#PeopleAndPlanetAI

product: Cloud - General; fullname: Alexandrina Garcia-Verdin; re_ty: Publish;