In continuation to the RAISE summit 2020, MeitY and NeGD successfully conducted another insightful session, AI Pe Charcha, on the topic “Open Source AI to Solve Billion Scale Problems” on 4th March 2021.

 Mr. Abhieshek Singh, CEO MyGov; President & CEO NeGD; MD & CEO Digital India Corporation (DIC) at Govt of India started the session with the introduction of the speakers of the day. In his opening note, he clearly stated that AI is no longer a technology of the future; it is very well here in our lives, and its involvement is only going to grow for better. He admired the efforts various industries are putting into adopting AI for their own betterment and making the lives of consumers much easier. He also said, “the idea of starting AI Pe Charcha was to continue the discussions started in RAISE 2020 summit and primarily to demystify AI.”

Mr. Vivek Raghavan, Chief Product Manager, and Biometric Architect at UIDAI. Trying to improve the UI around Aadhar authentication. He mentioned that there are high prospects that a citizen may enjoy most of the required services right on their smartphones. It requires proper identification, identity creation, and validation, speech recognition, fraud detection, Indic language translation.

He also elaborated on the scope and progress in face detection in India. The goal is to include facial recognition in every Aadhar related service, with the hardware currently supporting one crore authentication per day. The accuracy of these deep learning facial recognition systems is 99.4%.

This discussion was followed by a detailed demonstration of the Aadhar attendance application.

The next interesting project that grabbed the interest of the audience was the “Fingerprint Liveness Detection.” This robust AI model developed by UIDAI is tuned to differentiate between live and non-live fingerprints (non-live materials such as Fevicol or silicon and more) to prevent misuse. It is quite scalable and 99.1% accurate.

An interesting point that he mentioned was that India has a great potential to do wonders in the field of AI. India is very data-rich given our population, and data is all AI needs which gives us an advantage over others.

IIT Hyderabad and UIDAI are also extensively working on the deduplication project using deep learning. Another critical process is document validation & OCR with minimal manual intervention to progress towards a more digital India.

Mr. Vivek Raghavan then discussed another project, “ML based demographic deduplication,” which is helpful with kids whose biometrics is difficult. A logistic regression model trained on labelled data is used to calculate similarity based on the score on which suspected duplicates are filtered out.

This wonderful amalgamation of AI has huge potential in age and gender prediction too. Biometrics based on fingerprint, iris, and face make the backbone of the Aadhar system.

Mr. Vivek Raghavan is also heading another AI-based project with EkStop Foundation, an NGO, which is leveraging AI in Indian languages for public benefitting initiatives. The project is solely based on open-source frameworks. This is an end-to-end speech recognition system that aims to cover as many Indian languages for speech-to-text conversion.

The session was also quite open to questions and answers with a large audience from all genres, getting clarity and answers to many of their questions. The session also has elaborate demonstrations to bring clarity to people around the processing of the associated software.