To give more insight into the various initiatives taken by the esteemed institution, IIIT-B Director Professor Sadagopan talked to Financial Express Online’s Bulbul Dhawan.
Some of these projects have been envisioned against a backdrop of the global coronavirus pandemic.
IIIT-B upcoming projects: The International Institute of Information Technology, Bengaluru (IIIT-B) announced 11 visionary projects at the Bengaluru Tech Summit this year. These spanned across the EdTech, MedTech and AgriTech sectors, hoping to change the way these fields function. Some of these projects have been envisioned against a backdrop of the global coronavirus pandemic, while others are hoping to address long-standing problems in these fields. To give more insight into these initiatives taken by the esteemed institution, IIIT-B Director Professor Sadagopan talked to Financial Express Online’s Bulbul Dhawan. Here are edited excerpts from the interaction:
Drones seemed to be a big focus at the Bengaluru Tech Summit discussion. Could you tell us more about their utility in various fields?
Drones are autonomous vehicles, which we primarily look at as a new wave of AI and robots. Historically they were viewed to transport things and people, and therefore, the applications for drones as viewed by the western world were healthcare and logistics. For the Indian context, we thought of other applications, like precision agriculture. The biggest problem with agriculture in India are the very small farms. And we wondered if for them also, we could undertake image processing to assess the health of the crops and possibility of any disease and if we could take the action and spray pesticides to contain diseases among these crops.
With remote sensing, India has played a leading role, but unfortunately, it doesn’t work for agriculture. In this field, the images need to be processed really quickly so that action can be taken within a day or two to protect the crops.
Another application is in the area of traffic accidents. India has the largest number of road accidents in the world, and about 1.25 lakh deaths take place due to road accidents every year in the country. Mostly, deaths occur because the blood flow is not contained in time. So we looked at drones that would reach the accident spots reasonably fast and contain the blood loss, so that the country would be in a position to save many more lives.
In a way, the IIIT-B is focusing on drones in the Indian context for the larger social problems unique to the country.
Do you think drones are disruptive innovation?
I think drones are disruptive innovations in a few ways. One is, historically, avionics and aeronautics were seen as technologies that would need an airport and large infrastructure, and which would be for a relatively smaller number of people. Looking at the earlier days of automotive, driven by a driver equipped with special skills, needing to go to the service very frequently, we see that it took some time before it reached a stage where most people can drive and have spaces at home to park the cars. Essentially, we reached a stage of mass deployment. Similarly, drones can also be looked at planes for the common people in a way.
Another thing is that trucks are the mainstays of logistics. Due to their size and the fact that an individual usually does not require truck-worth of anything, they usually start and end at a particular place where an individual has to go and drop their things or arrange for that to be done. At this common point, the truck companies pack all the items of different individuals in containers, load them, transport them, unpack them and then distribute them. Instead of that, drones look at making these processes point-to-point, a facility currently available to only a handful of people.
To me personally, a more satisfying outcome is that it would lead to a new-age transportation of goods from person to person, especially over a short distance. If you look within a city, a lot of small things keep happening, like the transportation of documents or small items for which drones can be much more efficient.
The reason why we couldn’t think of drones for all these years was that their coordination is much harder. But now, we have the technology in which we are used to routing millions of things like YouTube videos, which has made it easier to apply that type of technology to control drones from a report point.
How is AgriTech growing and what is India’s role in it?
We have not looked at agriculture for all these years, and even for many youngsters, it is not cool. Then we spent some time with senior agricultural scientists, who told us about issues, challenges and opportunities in agriculture and it was mind-boggling. There are so many opportunities for science and technology in agriculture, and every aspect of it, be it weather forecasting, the right harvesting technologies, the supply chain and the constant monitoring.
Three of the IIIT-B alumni are running a start-up and they are working on dramatically increasing the yield and the margins that agriculturists get. The amount of wastage of fruits and vegetables in India is large due to the disconnect between the supply and the demand and they are trying to work on somehow bridging this gap with the supply chain technology brought to the farmer level. With that, the farmer would be able to directly connect with the customer, in a group of customers if not individually.
Do you think the step taken by your former students is a step ahead of the Centre’s e-NAM initiative?
Connecting farmers to customers is one part of the project. You have to make the farmer take the value out of it. The internet was kind of available in the early 1990s, but it took a while before Google came and people could search the internet. Similarly, Amazon had to come so that people could switch to online purchasing and Facebook had to come for social media. Building the internet was not sufficient.
This looks at letting farmers plan ahead for three years or five years, much like corporates do, and if they do it, can they also access bank credit not day-to-day but also over a period of time, and can forward trading can be undertaken in this as well.
What stages of development are these 11 projects in?
These projects are still in development and I think it would be unfair for us to claim that we have solved the problems, and we are just starting. We hope that in the next couple of years we can start delivering.
How did the team think of developing a platform like BeIYo and what need does it hope to cater differently from the currently available infrastructure?
In the current situation/infrastructure, all the medical records are digitally created by healthcare facilities in isolation. When the need arises, the concerned individual/patient can access the same in physical or digital form. This means that there is no system to verify the authenticity of such records in real-time. With this being the current scenario, numerous cases of counterfeit COVID-19 test results that’s being produced by individuals for travel and other purposes have become prevalent.
In an effort to mend the current situation, the IIITB team has launched, BeIYo – India’s first COVID-19 blockchain platform jointly developed by YoSync (a company incubated at IIIT-B) and a Malaysian based blockchain startup BelfricsBT in joint research with IIIT-B and funded by Mphasis.
Before developing BelYo, the team also created an education blockchain DAAP (Decentralised application) where educational institutes can create graduation certificates and related materials. This is jointly owned by the student who can share his/her credentials to prospective employers and can be verified in real-time. The product, CrediBel is currently being tested at IIIT-B.
Coincidentally, on 15th August, 2020, PM Narendra Modi launched National Digital Health Mission (NDHM) and as per the draft policy document, BelYo has already developed 2/4 major modules that are mentioned. The IIIT-B team is confident about being one of the service providers of NDHM in the future and has applied for the NDHM sandbox challenge.
How did the team work on coding the Margadarshi bot to ask questions that are of the level of an IAS exam, given the difficulty of IAS as well as the subjectivity of interviews?
IIITB connected with the India-for IAS Academy for the questions and shared the same with the BOT. The experts shared their insights and thoughts and their experience in the field over the years. IIITB’s technology allows follow up question generation based on the candidate’s answer, which is based on Deep Neural Networks, specifically transformer based models. In the future, there is hope to enhance this further with the data from the UPSC domain as well. IIITB is also in preparation to make the Agents more expressive and genuine.
Could you explain more about the Bar Chart Digitise project?
This project aims to achieve an automated reading of images of charts, e.g. bar charts, and scatter plots. These charts are available in image format predominantly, say in research articles, textbooks, and newsprint media. We are currently using image processing techniques for extracting the data that has been used for plotting the charts. We use machine learning models (neural network models) for chart type classification and text extraction. We also use optical character recognition for detecting text. Our future goal is to identify appropriate machine learning algorithms or models to improve our results in data extraction. We have tested our method for several datasets of chart images, and are currently building a tool which can take any user input of a chart image from the web or scanned image, and extract the data table, re-generate the chart at higher resolution, redesign the chart, etc. We have currently worked on digitizing images of bar charts, including seven sub-types. Solving the problem space for popularly used sub-types of bar charts, e.g. stacked bars, grouped bars, etc. is our novel contribution.
Charts are one of the simplest visual representations of data. There are several applications where the completely automated reading of charts will help, say in identifying key statistical trends in several charts generated for large datasets, for several runs of descriptive statistical analysis, etc. One of the key areas where computer vision of chart images can help is in promoting STEM education for the visually impaired, and in incorporating information from charts in documents for screen readers. However, the state-of-the-art is still in semi-automated solutions, owing to the vast design space in formatting charts used in the plotting library, the high amount of accuracy needed in reading charts, and the sparsity of pixel information in the images themselves. The sparsity in pixel information helps in increasing data-ink ratio and is useful for human reading, but is counter-productive for an automated reading. Given these challenges, this area is still ripe for AI/ML solutions. Hence, our interest in the problem statement of chart image digitization.