Identifying Urban Makeshift Communities using satellite imagery and geo-coded data
Arnab Biswas
@arnabbiswas
The aim of this talk is to provide a comprehensive description of the experimentation & explorations done by DataKind-Bangalore to identify non-permanent urban poor communities in Bangalore using Machine Learning (Transfer learning) with satellite imagery and geo-coded data for Pollinate Energy.
Pollinate Energy is a social business bringing life changing products such as solar lanterns, improved cook stoves, solar fans and water filters to urban makeshift communities. They identify communities by manually scouring through google maps and then visiting door-to-door. This is an expensive and time consuming process. Volunteers at DataKind Bangalore are working to help Pollinate Energy reduce man-hours and resources by suggesting urban makeshift communities using recent Computer Vision & Deep Learning Techniques.
Outline
This session will be organized as:
- Problem Description
- Data Gathering
- Challenges with data
- Various Modeling Approaches
- Satellite Imaging Approach
- Data Sources
- Computer Vision Approach (No labeled data)
- Supervised Learning Approach
- Transfer Learning Approach
- Geo-Coded Data based Approach
- Data Sources
- Undersampling Approach
- Stratified sampling Approach
- Satellite Imaging Approach
- Closing Thoughts/Learning
Requirements
Basic understanding of Machine Learning and Computer Vision terminology
Speaker bio
I volunteer with DataKind Bangalore. Professionally, I am working at Infrrd AI on Machine Learning.
Links
- DataKind Bangalore: http://www.datakind.org/chapters/datakind-blr
- Pollinate Energy: https://pollinateenergy.org
{{ errorMsg }}