Celestini Program India 2017-2019

If you would like to solve an important social problem while working with the latest technology and learning from incredible mentors, we have a program for you. You can win cash prizes and meet technical pioneers in the internet and communications. This program, called Celestini Program , has caught the interest of undergrads from both India and Uganda

Our competition goal focuses on solving an important socio-economic problem (such as road safety, air pollution, water quality) using machine learning and data analytics on a network of IoT or mobile devices where the data is collected from sensors, such as video cameras, air/water quality sensors, and used in real-time to make inferences, update models, and generate alerts. We solicit student teams to participate in this program led by Dr. Aakanksha Chowdhery (Google AI and 2012 Marconi Society Young Scholar) and Prof. Brejesh Lall (IIT Delhi) as directors.

The Marconi Society is a foundation dedicated to benefiting mankind through scientific achievements in the internet and communications. Each year, we recognize a group of inspirational graduate researchers with the Young Scholar award.

Our Young Scholars are passionate about working with students in developing countries to help solve important social problems. This work, called the Celestini Program, has successfully improved the health of pregnant women in Uganda and expanded to India in 2017 in partnership with IIT Delhi.

Brief summary of 2018 competition

More than one hundred students expressed interest in working on the program during the summer and three student teams (comprising of eight students) were selected. They chose problem statements related to air pollution and road safety in New Delhi were selected.

Air Pollution:Preserving the air quality is a critical challenge in the industrial and urban areas of many emerging economies. According to the World Health Organisation (WHO) global air pollution database released in Geneva, India has 14 out of the 15 most polluted cities in the world in terms of PM 2.5 concentrations. One potential solution is to increase awareness of the problem by enabling users to understand and track the level of air pollution, providing the basis for taking effective pollution control measures.

Clair: A student team developed a temporal forecasting solution based on the historical data reported by Central Pollution Control board to predict the real-time and fine-grained air quality information in five locations of Delhi.

    Prototype:
  • Air quality prediction of major air pollutants over the next 24 hours.
  • Tracks daily and seasonal variations of the major pollutants
  • Identify potential sources
    Results:
  • Developed an advanced machine learning model called CLair using LSTM techniques.
  • Deployed their solution on Google Cloud Platform to automatically generate predictions Tracks daily and seasonal variations of the major pollutants
  • Predictions updated every few hours on this website for five Delhi locations
  • Demo youtube video

Air Cognizer: Temporal forecasting model to predict the air pollution levels in locations where the Delhi Government provides air quality data hourly, but this is limited to specific locations. A scalable approach requires crowdsourcing where we use inputs from the entire population via smartphone applications widely used by Delhi residents that was developed by the second team.


    Prototype:
  • Android smartphone application allows users to upload an input image of the sky horizon taken from their smartphone camera..
  • The app predicts air quality particulate matter indicator, PM2.5 concentration, with an error less than 5% based on the certain features of the sky, such as how blue it is.
    Results:
  • Developed a machine learning model with image preprocessing using Tensorflow Lite to generate estimates by combining a pre-training machine learning model with a model trained online for each location based on a.
  • Image data preprocessed collected from different smartphone cameras so that the machine learning model works accurately
  • Deployed on the smartphone with Tensorflow Lite to enable a low-latency real-time prediction experience
  • Demo youtube video provides a sample of the Android application that the has launched in Google Play Store.

Vehicle-to-vehicle communication for road Safety: Another challenging problem that the student teams worked on was road safety. Over 200,000 people in India lost their lives in road accidents in 2015. Traffic accidents are the top cause of death for people aged 19-25 and often such traffic accidents involve multiple vehicles at high speed. This year, a student team worked on prototype solution that allows multiple vehicles to talk to one another at low latencies (tens of milliseconds) to send real time alerts about possible impending collisions to drivers behind them to prevent chain-reaction car accidents.


    Prototype:
  • System leverages computer vision to classify a given scenario as one that may result in collision.
  • Vehicle to vehicle communications to broadcast alerts.
    Results:
  • Each vehicle, acting as a node, broadcasts information related to its speed, location etc. and the other nodes receive and process this information based on the degree of relevance that the message holds. Developed a machine learning model with image preprocessing using Tensorflow Lite to generate estimates by combining a pre-training machine learning model with a model trained online for each location based on a.
  • The solution approach was designed over Xbee radios as a low-latency solution (~30-40ms) available off-the shelf at low cost.
  • Demo video and website link showcases their work during the project.

2018 Award Ceremony: The concluding ceremony of this year will held on November 1 at IIT Delhi where director of Marconi Society, Prof. Andrea Goldsmith (Stephen Harris Professor of Electrical Engineering, Stanford University), will give an inaugural address. The ceremony will be attended by the IIT Delhi Dean Alumni Affairs & International Programmes, faculty members from EE and CS, as well as industry partners. The winning team will be awarded a cash prize of $1500.

Brief summary of 2017 competition

30 students participated in Phase I and 6 students in Phase II. The team of 6 students in Phase II created a testbed in IIT Delhi experimenting with collaborative driver assistance to prevent collisions with pedestrians and other vehicles; designed using cameras connected to Raspberry Pis that communicate to cloud via mobile phone app. Their poster presentation was accepted at top-tier ACM MobiCom conference. This technology is being considered as a use-case for 5G pilot deployment in IIT Delhi next year. The student team is motivated to continue graduate school after their undergraduate work.

2017 Award Ceremony: Prize was awarded to the winning team Drizy on Nov 10 and was attended by Vice Chairman of Marconi Society Robert Tkach. The agenda and flyer for the award ceremony are as follows.

The video of the winning entry is on this link.

How it Worked in 2017 Our program focused on increasing road safety by using analytics from network of cameras mounted on vehicle dashboards.

Phase I (Feb 15 - Jun 1): Ten teams participated in Phase I from all over India. Each student team prototyped video analytic algorithms on Raspberry Pi 3 Model B to process video data of cameras mounted on vehicle dashboards. The video analytics algorithms analyzed real-time from video of camera facing front of the vehicle: detect pedestrians and their trajectory, infer position or lane of the vehicle on the road, and the density of vehicles in front of the vehicle where camera is mounted. Each team trained their vision models, but the model accuracy will be tested on dataset provided from Delhi highways.

Phase II (Jun 1 - Sep 15): The top team from Phase I entered phase II. In phase II, the team worked at IIT Delhi to build a testbed where a network of cameras update information and receive alerts from backend servers with a visualization API to detect vehicle-to-pedestrian collisions and vehicle-to-vehicle collisions.

Details for 2019 Competition

Each participating team must be comprised of two-four undergraduate students enrolled at an Indian engineering institution. The team can also involve a faculty mentor from their institution. The team must arrange equipment at their own cost for Phase I. To participate in 2019, please submit your team details by early March, 2019. For questions, please email celestiniprizeindia-AT-gmail.com.

Phase I: We will give a take-home exam (comprising of coding+open ended design questions) on a spring weekend. Two participants from each team must take the exam for their team to qualify. Each team will submit a document on their proposed approach to solving problems around theme given by organizers.

Phase II: The selected teams will intern at IIT Delhi for three months between May 15-Sep 1. Each team will comprise 2-4 members and each student team member will receive a stipend of $200 for the internship. The top team will receive Phase II prize of $1500 and their entry will compete in Celestini Prize at International level.

Prizes

The prizes are sponsored by Marconi Society and Google.

Stipend for Phase II: The teams entering Phase II are awarded cash stipend to work at IIT Delhi: $200 per student

Prizes for Phase II: The winning team with best working prototype will win a $1500 cash prize and will compete for Celestini prize with a team from Uganda.