I am passionate about building intelligence. I am interested in multimodal machine learning at scale.
Data Scientist @ Using deep learning techniques to solve practical problems. From November 2015 to Present (2 months) Mumbai Area, IndiaComputer Vision Engineer @ Accomplished a wide variety of projects ranging from machine learning and web scraping in C++ and Python to writing backend server infrastructure in PHP. Experience in feature engineering and also writing production level optimized code.
Summary of projects:
1. Feature engineering - Developed new vision features for our product Vidalife in order to quantify the interestingness of customer events at home. (Check out http://vidalife.info)
2. AWS Cloud setup - Established the entire AWS cloud infrastructure for our product requirements. Worked with AWS tools such as S3, SQS, EC2 and CloudWatch.
3. Web scraping - Wrote scripts for scraping e-commerce sites to identify the main strengths and weaknesses of competitor products as well as identifying potential customers.
4. Blink detection - Eye blink detection using the Random Forest learning algorithm and Histogram of Oriented Gradients (HOG) as features. Collected a balanced dataset of 300,000 samples
and achieved an accuracy of 92% on 50,000 testing samples.
5. Wrote efficient, production level C code for the embedded board on our product. That is, ported some of our research code to efficient production level code.
6. Web development - Extensive work in Javascript and PHP while developing various proof of concept demos like depth sensitive face tracking Twitter demo. Additionally, two tools in PHP for better managing our cloud resources.
7. Product setup experience - Developed the device/server side Bluetooth 4.0 (Bluetooth Low Energy) communication service for interacting with iPhone using Node.js (javascript) and Bash.
8. Feature engineering - Currently, developing new audio-based features for further improving the accuracy of our algorithm for quantifying the interestingness of customer events at home. From June 2014 to Present (1 year 7 months) Bangalore Area, IndiaThesis Student @ Developed an algorithm for super pixel clustering using kernel density estimation. Posed image segmentation problem as that of finding optimal search space. We find optimal segmentation by using a combination of under-segmented and over-segmentation scale.
- Developed a new method for determining sigma parameter for gaussian kernel density estimation.
- Further applications include semantic segmentation, object detection.
This work was published in ICVGIP 2014. See publications below for more details.
In addition, worked on a simple and efficient algorithm for determining object saliency. Some Results: http://val.serc.iisc.ernet.in/SaliencyResults/ From January 2014 to May 2014 (5 months) Summer Intern (Google Summer of Code 2013) @ Inpainting can be viewed as a quite powerful tool for removing and/or repairing small parts of the image. It can be used to remove some small unwanted parts in an image and at the same time "fill-in" some missing parts.
Image Inpainting using Fast Marching Method developed by A. Telea is one of the most popular algorithms due to its speed and decent results with thin regions. Though, the restoration obtained suffers from blurring and fails for restoring large regions. Another algorithm developed by A. Criminisi for Texture and Structural Synthesis using Exemplar method though slower in comparison with Teleas's method provides brilliant (almost distortion free) results for texture synthesis and reconstructing linear structures. In this GSoC I have, implemented three algorithms including the above two for the purpose of Image Inpainting and offers users a choice between speed (less than 2 seconds - FMM) and very accurate results (Criminisi's).
Instead of implementing the 3 modules initially proposed in my proposal we decided to implement 2 algorithms in 1 module and skip a module completely. The initial project proposal can be viewed here: www.google-melange.com/gsoc/proposal/review/google/gsoc2013/chintak/1
And you can follow the blog on analysis and implementtaion of the algorithms here: http://chintaksheth.com/
Scikit-image Website: http://scikit-image.org/ From May 2013 to September 2013 (5 months) Co-Founder and Student Coordinator @ We started this department as a way of contributing back to the campus and helping create a technical community. A platform for passing knowledge from seniors (excelling in a field outside the scope of our curriculum) to juniors.
We explore topics like Web design and development, C#, Python, Image Processing, Computer Vision, Robotics, 3D modelling using CATIA and ANSYS which are not currently taught as part of our curriculum. The idea is to supplement the BITS curriculum and provide assistance to students wanting to explore different fields in technology.
Employed a paid model, which not only provides token fees to senior students teaching the course but also serves as an incubator to provide monetary support for technical projects on campus. This semester we funded 8 projects! The courses are designed to be more hands-on and contain hackathons and project submissions as evaluation components. For more information: https://www.facebook.com/groups/CTE.BITSGoa/
Website: http://cte.bits-goa.com From November 2012 to May 2013 (7 months) Roboficial Panel Coordinator @ Quark is the annual Techno-managerial festival of BITS Pilani Goa campus.
As the Roboficial Panel Coordinator my duty is to supervise all the events related to robotics at Quark 2013 which include Robokombat (our flagship event), all the way from their design inception to its execution. From June 2012 to February 2013 (9 months) Summer Intern @ I am currently working on designing a software for the company to efficiently manage the salary hand-out of its employees, to provide transparency to the whole process.
Such a software already exists in the company, but we have been assigned this project to serve as an alternative to the monopoly of the person in-charge. From May 2012 to July 2012 (3 months) Summer Intern and Trainee @ From May 2012 to June 2012 (2 months) New Delhi Area, IndiaMember @ Procurement of sponsorship for the International Conference on Interactive Systems 2012, organized by the IEEE Student Branch at BITS-Pilani Goa campus.
Our main objective is to provide a platform for reputed people in the filed of Augmented Reality, Interactive Robotics, Human Computer Interaction etc, to interact and exchange their knowledge to solve the problems faced by us today. From April 2011 to April 2012 (1 year 1 month) Core Member @ From July 2011 to January 2012 (7 months) Member @ Waves is the annual Cultural festival of BITS-Pilani Goa Campus.
The objective of the Waves Research Team is to evaluate the performance of the previous Waves and to come up with innovative and creative solutions in order to solve the failure. Also to come up with ideas to make the up-coming Waves a greater success. From May 2011 to November 2011 (7 months) Member @ The Department of Finance and Asset Management of BITS Goa is responsible for handling the money and the assets for all the three festivals of BITS-Pilani Goa campus, viz. Waves (Cultural festival), Quark (Technical festival) and Spree (Sports festival). It is concerned with the sensible investment of the assets available in order to make the above stated festivals a success. From August 2010 to April 2011 (9 months)
B.E. (Hons.), Electrical and Electronic Engineering @ Birla Institute of Technology and Science From 2010 to 2014 Chintak Sheth is skilled in: C, Amazon Web Services (AWS), Matlab, Python, Programming, Image Processing, Event Management, Verilog, Research, Robotics, Computer Vision, Public Relations, Entrepreneurship, AVR, Cadence Virtuoso
Websites:
http://chintaksheth.com