1) Legend of Tana (2-d Game)
I’m glad to make my first developed game called as the ‘Legend of Tana’ as a freeware. I developed this game way back in 2007, and naturally, it was compatible with only Windows XP and the previous versions of Windows. I later had compatibility issues which are now been resolved. Now the game is compatible with Windows 8/10! My game also won the weekly edition of Intel Game Demo Contest held in 2007.
Working to launch the 3d version of the game soon!
Comment your feedback on this game or report any bug which you experience.
Download Game at
2) Web Scraper
I wrote a web scraper to fetch e-book pricing information from e-commerce portals. This is written in Python and built upon Scrapy(A Py Framework).
3) Binary Classifier for Text Images
This is a binary classifier built with machine learning techniques to segregate English from non-English text images. This is to reduce the noise, given to the classifier during the Pre-OCR stage. This is again a Py implementation with LibSVM module written in C++.
4) Efficient Binarization and Natural Image Segmentation
I implemented an algorithm which gives 20% speedup than the Sauvola Binarization algorithm, and was used to extract natural images from text document images in the pre-processing stage of OCR. This was developed in partnership with Amazon India Development Centre.
5) Video Script Identification using GPU
I wrote some modules in python, C++, and CUDA to detect the language/script from the videos. This is done as preprocessing steps to recognize scripts of different languages which is later use to feed the OCR to generate annotations for Video Search Indexing. This reduces the overheads in removing OCR-noise during pre-processing.
6) AirTeach
AirTeach is a web-based teaching module, which is packed with essential utilities for a modern-day teacher. Teachers can use their mobiles to navigate through slides, just by shake/any other gesture. They can also rotate/peel off a 3d model, just by rotating their smartphones. Further, they can access the desktop mouse for highlighting content, just by moving across their smartphone screen!
Github Link: https://github.com/kaushikmit/universal_handheld_kinect
For those of you who are curious, read this!
7) Semantic Error Detection in Books
Developed a N-gram based ML model to spot out the semantic inconsistencies in a book. This could identify contextual errors, and report them instantly. This helps in improving the book quality automatically over time.
8) Sketch Suggestions for Doodle
Developed a multi-class classifier for classifying image sketches based on character/glyph resemblances. This helps in auto-suggesting clip-arts/sketches to make instant doodles. Implemented the neural nets using Keras and Tensorflow.
9) Tamil Speech Recognition using CMU Sphinx
Developed pre-processing modules for Tamil Speech Recognition under the supervision of Dr.Madhan Karky
– Developed a web crawler to fetch Tamil content from websites
– Identified the morphological complexity of Tamil Language
– Constructed a word to syllable dictionary
– Created an efficient N-gram language model for Tamil
– Developed a system that can output phonemes for a given speech sentence using CMU Sphinx
