You are here


Researchers Develop a Method to Estimate Headcount of Crowd Using Static Images

On February 10, 2013, 30 million people descended to Allahabad to bathe at the Sangam (the confluence of Yamuna and Ganga) on the occasion of Mauni Amavasya, the most auspicious day of the Maha Kumbh Mela. Disaster struck the Allahabad railway station when about 200 thousand passengers overcrowded the station and rushed to a foot over bridge to change platforms causing a stampede. The bridge collapsed due to the sudden pressure, killing 36 people and injuring at least 39. Can such disasters be avoided? Yes, say Prof. R. Venkatesh Babu and his teammates, Lokesh Boominathan and Srinivas SS Kruthiventi from the Video Analytics Lab at the Indian Institute of Science (IISc), Bengaluru. Using Neural Networks and Deep Learning models, they have developed a method to estimate the crowd density and the number of people in the crowd with just the image of the crowd.

A Google for Machines?

Lee Sedol, a renowned Go champion recently lost to AlphaGo, an Artificial Intelligence (AI) system developed by Google DeepMind. Experts believed such a system was five or even ten years away from now, owing to the complexity of game. Many regard the success of AlphaGo as an indicator of the pace in which AI - the field focusing on developing machines capable of intelligent behaviour -- is progressing. Dr. Partha Pratim Talukdar is pioneering contributions to this field with his Machine And Language Learning (MALL) Lab at the Indian Institute of Science, Bangalore.

Scientists at IISc working towards video surveillance

Scientists at the Indian Institute of Science, Bangalore, have developed an algorithm that helps monitor people’s movement under video surveillance systems. These systems, also called Close Circuit TV (CCTV), have become ubiquitous and generate enormous amounts of video data. Analysing this data manually to detect abnormal behaviours and possible dangers is a huge task. A team of computer scientists led by Prof. R. Venkatesh Babu at the Computational and Data sciences Department are working actively to tackle this problem.

Building a Smarter India, the DREAM way

The recently launched 'Smart Cities Mission' is a bold initiative by the Government of India with the goal of upgrading 100 cities into “smart” cities. The mission aims to drive economic growth and improve our quality of life by using technology. 'Internet of Things' and 'Big Data' are today’s buzzwords promising us a life out of a sci-fi novel. But is it possible? The dreamers at the DREAM (Distributed Research on Emerging Applications and Machines) Lab at the Indian Institute of Science, Bangalore, are working on exactly that!

DREAM:Lab – Democratising Computing through the Cloud

The history of computers dates back to the invention of the Abacus - invented to help ancient merchants in trading cows efficiently and getting rid of calculating by hand. But the appealing history of machine-assisted human computation and modern computers originated only around sixty years ago. Since then, computers have transformed from room-sized mega-boxes to desktops to laptops to our mobile phones that fit in the palm of our hand. The rapid evolution of technology with wearable computers, embedded chips and smart appliances is aiming towards providing a better quality of life to humankind. This evolution is proceeding to make the machines talk to each other with the much-hyped “Internet of Things” technology. On the other hand, advancements of machines to aid scientific discoveries by building smarter, faster and more capable computers has led to the inventions of “Supercomputers”.

An online tool to study Hydrogen Bonds with precision

Scientists at the Indian Institute of Science, Bangalore, have developed the world’s first Hydrogen Bonds Computing Server (HBCS) which computes the hydrogen-bond interactions with their precision. The eleven-member research team was lead by Prof. K. Sekar, Associate Professor at the Department of Computational and Data Sciences.

Expresso – A smart way to build Neural Networks

Google made headlines when its computer program powered by “deep neural networks” defeated a professional Go player. In the recent past, machine learning is catching all the attention and scientists in our backyard are contributing to it. Ravi Kiran Sarvadevabhatla, Jaley Dholakiya and Prof. R. Venkatesh Babu from the Video Analytics Lab at the Department of Computational and Data Sciences, IISc, have developed a user-friendly interface for designing, training and developing such “deep learning” neural networks. Their framework, called Expresso, is written in the Python programming language and is free.