Embedded Files
INTRODUCTION
INTRODUCTION
I completed my B. Tech. in Comp. Sc. & engg. from IIIT DM Kancheepuram in July 2019. After which I did a machine learning internship at IIT Bombay where I contributed in experimentation and development of various deep learning techniques for a form layout and text detection software stack.
I have also have research experience working as a research intern at IIT Madras and at IIITDM Kancheepuram.
You can find my CV here.
Education
Education
IIITDM - Kancheepuram
Bachelor of TechnologyIn Computer Science and EngineeringCumulative GPA: 8.18/10.00Sumbeam School Lahartara
Higher SecondaryScore: 93%Rajghat Besant School
MatriculationScore: 10.00/10.00Research Interests
Research Interests
My research interests lies in Computer Vision, Deep Learning, Machine Learning and Generative Models.
Projects
Projects
Hand filled form schema detection and OCR
Mentor: Pankaj Singh, Dept. of Comp. Sc. IIT Bombay linkedinHand filled form schema detection and OCR
Development of an annotation tool Form Labeller using Python + Tkinter. github
Trained a text detector yolo v2 model.
Worked on Form Schema detection, this was a three step pipeline, with box detection and OCR and then schema detection.
Developed novel form alignment digital image processing algorithm.
Obfuscated, binarized and licensed two projects for clients using pyarmor and pyinstaller.
Generative Adversarial Models for Fashion Design Inspiration (Bachelor thesis project)
Guide: Dr. V. Masilamani, Dept. of Comp. Sc., IIITDM. siteGenerative Adversarial Models for Fashion Design Inspiration (Bachelor thesis project)
Trained a deep convolutional generative adversarial network on a subset DeepFashion dataset to generate images of women’s T-Shirts which can be used as inspiration for new dresses by fashion designers.
Two neural networks were trained, one generative the other discriminative, in an adversarial fashion.
Used Python and Pytorch to demonstrate results.
Face landmark detection using convolutional neural networks
Guide: Dr. V. Masilamani, Dept. of Comp. Sc., IIITDM. siteFace landmark detection using convolutional neural networks
Trained a neural network which can identify 64 landmarks on human faces.
Used transfer learning to train a ResNet on a limited dataset in Python using PyTorch library.
Use of Instagram photographs to identify influential fashion bloggers by finding similarities in photos posted by users and their followers
Guide: Dr. Nargis Pervin, DoMS, IIT Madras siteUse of Instagram photographs to identify influential fashion bloggers by finding similarities in photos posted by users and their followers
Big data web scraping using Python and web-based automation tool Selenium.
Developed a code for Apparel classification in images employing Convolutional Neural Networks using Pytorch in Python, 84% top-5 accuracy obtained.
Used Python and Pytorch to demonstrate results
Structural and algorithmic properties of Max chordality 4 and maximum degree 3 graphs
Guide: Dr. N. Sadagopan. HoD, Dept. of Comp. Sc. IIITDM. siteStructural and algorithmic properties of Max chordality 4 and maximum degree 3 graphs
Defined structural properties and graph generation algorithms.
Found polynomial time algorithms for classical problems like vertex cover, dominating set, feedback vertex set for the graph subclass - maximum degree 4 max chordality 3
Form Labeller
Form Labeller
This tool helps in modifying/creating existing labels/annotations of images.
GAN Implementations
GAN Implementations
Real and Generated Images using Deep Convolutional Generative Adversarial Networks.
Another implementation of GAN on faces dataset can be found on My Website(May take a little while to load).
Hobbies
Hobbies
I like to play football and play the ukulele while my models are training. I also like tinkering with raspberry pi and making retro gaming consoles.
External Links
External Links
Contact
Contact
devershigpt6 [at] gmail [dot] com
Google Sites
Report abuse