Projects

Filter Projects: [Computer Vision] [Deep Learning] [Machine Learning] [Software Development]

	Kaggle Right Whale Recognition
	Convolutional Neural Network model to recognize individual Right Whales
	We use several different methods to recognize individual Right Whales: fine-tuning a pre-trained CNN model, and reusing those models as feature extractors to fed into an SVM.
	Toolstack: C++, Python, Caffe
	Computer Vision Deep Learning Machine Learning Image Classification Fine-Grained Recognition
	PDF

	Bird and Squirrel Alert System
	System to recognize birds and squirrels on a bird-feeder from video
	We developed a system to automatically recognize and locate birds and squirrels on a bird-feeder in real-time from continous video input. We use optical flow, motion detection, and tracking.
	Toolstack: C++, OpenCV
	Computer Vision Object Recognition Object Localization
	Video

	Human Recognition in Infrared Images
	Convolutional Neural Network model to recognize humans in infrared spectrum
	We manually collect human and non-human infrared images using Seek Thermal imaging camera and manually label/annotate them. We fine-tune pre-trained CNN models and use them as feature extractors to train SVMs.
	Toolstack: C++, Python, Caffe
	Computer Vision Deep Learning Machine Learning Image Classification

	Kaggle Microsoft Malware Classification Challenge
	Machine learning model to classify 0.5 terabytes of malware programs
	We trained a machine learning model using XGBoost, scikit-learn to classify malware source code into 9 classes. Extracted features based on byte 4-grams frequency and instruction count.
	Toolstack: C++, Python, XGBoost, scikit-learn
	Machine Learning Classification

	First-person Scene Classification
	A model to recognize place, activities, and objects in first-person images
	We manually collect thousands of first-person photographs using Narrative Clip wearable camera and then we create a hierarchical structure of 17 classes in 3 main categories: places, activities and objects. We do multi-label classification using Caffe and scikit-learn's OneVsRestClassifier.
	Toolstack: C++, Python, Caffe, scikit-learn
	Computer Vision Deep Learning Machine Learning Image Classification Egocentric Vision

	Optical Music Recognition
	A system for recognizing musical notations from sheet music
	Implemented a system to recognize musical notations by using template matching, music notation boundary detection, and staves detection using Hough transform.
	Toolstack: C++, CImg
	Computer Vision Image Processing Machine Learning Object Recognition

< / >	BitTorrent Client
	An implementation of BitTorrent Protocol Specification
	- Implemented an object oriented multi-user BitTorrent client in C++ based on BitTorrent protocol specification, TCP/IP and network programming concepts. - Supported N-seeders and N-leechers simultaneously and validated file integrity using SHA1 hash.
	Toolstack: C++, Wireshark
	Object Oriented Programming Software Development Network Programming TCP/IP

< / >	Port Scanner
	Network security app to scan for open ports in remote hosts
	- Built a multi-threaded Port Scanner in C++ to scan for open ports in remote servers for security checks using raw sockets and various TCP/UDP port scanning techniques like SYN, UDP, ACK, and FIN scans. - Identified service version of processes running on remote server by parsing the response of port scans.
	Toolstack: C++, Wireshark
	Object Oriented Programming Software Development Network Programming TCP/IP

	n-Puzzle Solver
	A Java application to solve and evaluate n-puzzle
	- Developed an object-oriented, multi-threaded 15-puzzle solver (scalable to n-puzzle) using Java and parallel programming. - Remodeled the puzzle as a graph search problem and implemented and compared various search techniques like A*, Depth-first and Breadth-first.
	Toolstack: Java
	Object Oriented Programming Software Development Parallel Programming
	PDF