AI Student Ambassador Sumedh Pendurkar: Applying Pattern Recognition and Computer Vision for Text and Photo Analysis

Bringing Images and Scripts to Life Using Trained Models and SVM

Intel® Student Ambassador Sumedh Pendurkar gives an in-depth look at his work in computer vision and pattern recognition for his optical character recognition (OCR) and image resolution projects as well as his involvement in launching a satellite.  The Intel® Student Ambassador Program was created to work collaboratively with students at innovative schools and universities doing great work in the machine learning and artificial intelligence (AI) space.    

Tell us about your background and what got you started in technology.

 As a child, I was into computers and technology.  I used to look for upgrades in computer hardware and software, as I found it interesting.  I continued this passion of computer technology and am currently in my final year at College of Engineering, Pune (COEP) pursuing my Bachelor of Technology in Computer Engineering.  

Working on computer vision and pattern recognition problems is important to me.  I’m currently working on two projects:  OCR for Devanagari Scripts and Image Super Resolution.  Apart from that, I work with COEP initiative, CSAT, a Satellite Team that recently launched a satellite.  My work on the team focuses on designing communication protocols between various controllers and peripherals.

Tell us more about your projects.

One project I’m working on is OCR for Devanagari Scripts aims to pull text from pictures of printed documents.  I have trained Support Vector Machines (SVM) with linear kernel to classify the characters.  My final goal here is to segment words into characters.  The second project is Image Super Resolution, where the task is to generate a high resolution image from a low resolution image.  

Our satellite team launched our first satellite named "Swayam" on June 22, 2018.  The main objective was to establish point-to-point communication between the satellite and ground-station.  The second objective was to demonstrate passive altitude stabilization, and ours is the first Indian satellite to do so.  Currently, we are working on the second satellite mission whose main purpose is to justify the use of solar sails for space travel maneuvering from low to high earth orbit to collect useful scientific data.

You recently hosted a lab, can you tell us more about that experience?

The main motive behind hosting the lab was to introduce college students to the field of AI and machine learning to inspire them to start working in it.  Intel® Software Innovator Siddhant Agarwal started the lab with an orientation explaining basic algorithms and demonstrating useful applications.  Later on, I demonstrated my project, and ended with giving students ideas on how to start a project.

Based on feedback from the lab, I am also planning to conduct a training workshop on the basics of machine learning to get people started by undertaking a small project or competition.  I expect this to last for 2-3 weeks, taking a few hours a week.  In individual interactions with students after the initial session they showed a strong interest in learning more about machine learning so I’m keen to offer more sessions.

Tell us about a technology challenge you’ve had to overcome in a project?

Most of the work that I like to do involves very deep neural networks and requires a lot of computational power.  This is the case for most problems that involves using images as the data.  Thanks to Intel® AI DevCloud, Intel® Distribution for Python* and Intel® Optimization for Tensorflow* I was able to speed up my work.  As an example of how much improvement I was able to see, I was recently training a small variant of a zero-shot super resolution (ZSSR) model and my CPU took 1586 seconds per epoch.  When I switched to using the Intel AI DevCloud the processing time was reduced to around 20 seconds per epoch – an improvement of approximately 78x. The time saved allowed me train more models and achieve better results.

How are you planning to leverage Artificial Intelligence or Deep Learning technologies in your work?

My goal is to use machine learning and deep learning models to solve problems and achieve better accuracy.  I would also like to solve real life problems to help people.  For example, Devanagari scripts are commonly used in India and Nepal and my OCR project would help those countries to get documents translated to text by clicking pictures and feeding them to the algorithm.

As a Student Ambassador, what are you looking forward to doing with Intel?

As a student, I’m always eager to learn new things.  I would like to improve my machine learning and deep learning skills using the Intel AI DevCloud. As an Ambassador, I would also like to conduct more sessions on machine learning and deep learning to expose more people in my community to the work being done in the field of AI.  I would also like to write more articles and publish on Intel's website to spread the knowledge and inspire others.

How can Intel help students like you succeed?

Often students give up on deep learning because they do not have enough resources to train their models.  Intel can help by providing the necessary software and hardware resources.  Apart from that, Intel can also conduct informational sessions about recent advancements in technology to further inspiration, innovation and creativity.

I believe that one should follow his passion and work on the things he or she likes.  This creates interest and excitement to try out your own ideas and the satisfaction of seeing them work.  Personally, I keep on reading new things, taking additional courses, and observing the code and output of algorithms that I find interesting.  This helps to keep my knowledge up-to-date and keep me inspired.

What impact on the world do you see AI having? And do you see yourself as part of it?

AI is developing day by day, finding applications in almost all fields including medicine, transportation, finance, entertainment, etc.  With the growing power of computation, it is becoming easier to develop algorithms which can make life easier by automating even non-trivial tasks.  I look forward to finding even more ways to use technology to help people.

Join the Intel® Student Ambassador Program for AI

Upon acceptance into the Intel® Student Ambassador Program for AI, Graduate and PhD students from top universities worldwide can access newly-optimized frameworks and technologies, hands-on training, and technical resources provided by the Intel® AI Academy. To be considered for the program, apply today.

Learn more about the Intel® Student Ambassador Program for AI or Intel® AI DevCloud. For inspiration or to start sharing your work, join our community collaboration platform Developer Mesh.

For more complete information about compiler optimizations, see our Optimization Notice.