research

Geospatial AI for cities and farms

Contents

Introduction

Do you know where the food you eat grows? Or do you know how much the river paths in your city have changed? Would you be interested in knowing, what are the most walkable paths in your city? Or where most accidents happen?

“Knowledge is power. Information is liberating. Education is the premise of progress, in every society, in every family.” - Kofi Annan

We live in a highly interconnected world. The electricity we use is often generated hundreds of kilometres away in a power plant, the food often travels thousands of kilometres before entering our homes and even the homes are built with cement manufactured thousands of kilometres away. In such interconnectedness, we often lose touch with how we’re using the environment in which we live. In this group, we would like to understand these interconnected aspects and promote mindful and sustainable usage of resources while ensuring economic prosperity for all.

Role of geospatial data

Satellites provide a picture of the entire world and in this picture, it captures a lot that is happening in the world; Farmers growing crops, cities building roads, people migrating places, rivers changing paths, water bodies changing color, forests being cut and much more. We want to capture these changes through our strength in Computer Vision, Machine Learning and Systems methods.

We would like to derive insights about different aspects of our lives, both in urban and rural settings. We would like to understand how cities function, where does the waste go, where does the water come from, how do people travel in the city, etc. In rural areas, we would like to understand the village’s land use, optimization of agricultural practices, development of new projects, etc. We would also develop an understanding of the natural resources, how the forests, rivers and lakes around us evolve. This information will empower decision makers, governments to make policies that keep the interconnectedness of the entire ecological system in mind and the public to take decisions regarding their lifestyles and vocations.

Current Focus

Our current projects are in:

  1. Urban Planning We build systems to inform choices in the urban landscape for efficiency with respect to urban climate and mobility. We’re building a perception stack to monitor urban resources - land use, roofs, trees, lakes, etc. in cities across India (and eventually, the world) based on remote sensing data. We are working on various aspects including heat island modeling, tree cover tracking, etc.

    We use these layers as an input to our climate and mobility engines which are used to assess climatic and mobility outcomes for different input scenarios. Our technical contribution is in learning semantic representations for these inputs, processes for easily annotating and scaling the resources to be tracked and finally, mechanisms to assess climate and mobility at the city level - primarily flow algorithms for mobility and physics inspired networks for climate.

  2. Agriculture In agriculture, our efforts are towards self-supervised learning algorithms for agro-climatic image understanding. Essentially, we believe that a generic image representation for dynamic scenes such as crops doesn’t exist in isolation from climate, soil and other topological variables. Therefore, we deviate from the traditional foundation model approach to develop small language models dedicated to agricultural tasks. Our recent work has shown promising results with this approach.

    We use satellite images and multispectral drones in conjuction with ground experiments. Most of our current work is with the Sugarcane crop. We’re working on computer vision models for stress and yield and systemic models for interactions with various elements such as soil health, weather, etc.

The image belows shows a representation of different components of our research:

Lab Philosophy

If you are interested in being part of the group, please read the lab philosophy. I hope this gives you an idea of what you could expect to work on as a part of the group.

Team

We call the people of the group GeoVisionaries :-)

Faculty

The group is lead by Anupam Sobti while a lot of the work is carried out in collaboration with domain experts such as:

PhD Students

  1. Varchita Lalwani
  2. Moti Rattan Gupta - moti dot gupta at plaksha.edu.in

Research Fellows

  1. Yogendra Kumar

Undergraduate Students

Alumni

Sponsors

  1. Indorama Ventures Center for Clean Energy, Plaksha University
  2. Center for Sustainable and Precision Agriculture, Plaksha University
  3. Plaksha University Startup Grant
  4. Google Research Cloud Credits Grant

Join Us

We currently don’t have open full time positions. Students may reach out as per information here.