Technology Expert: Artificial Intelligence, Geospatial technology tools
Brief Profile: Ayushi Mishra, a Forbes 30 Under 30 awardee, is a Biomedical Engineer and holds an MS Engineering Management from Johns Hopkins University. Her startup-DronaMaps specialises in large scale mapping with drones and extraction of geospatial analytics from it. Ayushi specialises in solution design at the intersection of design, technology, and business. During the Covid-19 Crisis, DronaMaps volunteered their platform to seven states across India acting as a command-and-control center. The deployment was inspired by Hopkins Covid-19 dashboards but ended up providing deeper analytics like patient tracking, healthcare infrastructure tracking, containment and hotspot analysis, predictions on spatial and epidemiological spread of the disease at a district level for almost 130 districts across India. Ayushi has been recognised by Vogue Magazine as one of eight women in STEM leading the battle against Covid-19 in India.
Journey as an entrepreneur: Ayushi started her first venture-- Marigold Health, while still in graduate school at Johns Hopkins University. It is a digital health platform with Artificial Intelligence to scale peer support and therapy. Marigold Health is currently funded by the National Institutes of Health and Rock Health foundation. In late 2017, she moved to India to work for DronaMaps, a company cofounded by her longtime friend and fellow Hopkins student-- Utkarsh Singh. DronaMaps has now mapped over 600 sqkm with over 100 villages and multiple Smart Cities.
Enterprise and its application: DronaMaps is a Command and Control Center solution to enable decision makers with a detailed overview of the geospatial layout of large scale assets with 3D drone maps. We use deep learning algorithms to extract 14 geospatial analytics used for sustainable development planning and precision agriculture. The platform integrates with surveys, images, telecom tracking (CDR/VLR), CCTV, sensors, and drone live feeds deployed on the ground. The analytics at the end provide actionable insights enabling resource and project management and monitoring. Dronamaps is a command-and-control center solution specializing in large scale 3D mapping with drones, we use deep learning to provide geospatial analytics for applications like Sustainable development planning, Smart cities and villages. Founded in 2016, DronaMaps was started with the vision of empowering decision makers with accurate geospatial data to make development planning and execution more robust. Our work found applications including, the Smart Village project with the University of California, Berkeley, Smart Cities, precision agriculture etc.
Technology Used and how it integrated into your product: We use a combination of legacy geospatial approaches with cutting edge drone technology. The perspective is to make it easier for decision makers to use geospatial data in day-to-day work without relying on specialised teams. The techniques like photogrammetric reconstruction from imagery captured through satellite into maps while ensuring the accuracy of the map, have all existed under the discipline of geospatial intelligence. Drone based mapping is an extension of this, which borrows from the broader discipline but builds on it, the swathes of areas captured in satellite imagery are huge, in the range of 200 sqkm and multiple points are matched, while in drone maps, more than thousands of points are matched, therefore, it is computationally intense. However, at the end of all this what you are left with is a gigantic image that needs to be mined for information. Our deep learning algorithms come into the picture at this point— to use the high-resolution maps captured with drones and identity 14 geospatial features required for development planning like plot boundaries, location of trees, roads, utilities like power lines etc. Initially, we were doing this in-house manually as clients pointed out the need for usability in drone maps, this manual tagging of 600 km2 of high-resolution maps, now serves as a master training database, that has been designed to account for multiple scenarios like type of drone used, resolution of maps, population and construction density of area mapped, even the time of data capture. Existing machine learning algorithms do not work so well on the complex geospatial terrains like India, which is why it helps to train the deep learning algorithms on Indian data and do so robustly. This cuts down the most manually intense aspect of mapping, after this point all you need to serve client needs is our platform that visualizes the city scale 3D maps and the analytics associated with it for a range of applications like infrastructure project tracking, Utility planning, etc. Our recent focus has been to create analytical filters which would enable prioritising of operational tasks based on the massive geospatial datasets.