Mila Luleva: Why Would a Bank Care About Satellite Images? (Spoiler: It's Carbon Credits) - MBM#41
Minds Behind Maps
The Challenges of Building an Accurate Model for a Training Set
Each plot that we have, each one of the 100,000 active plots has been collected by a person who took the polygon GPS and then verified by at least three quality checks on the way. So I can... With 98% accuracy guaranteed that those plots are the ones that we are monitoring and they belong to those farmers. The more and more we worked with this, the more and more I found out that actually the bigger issue is how you apply that model. It's so much more different and difficult to monitor to you cannot validate that. You're not estimating it or validation set, not on the application. Yeah.
00:00
Transcript
Play full episode
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
