Abhinav Kumar | अभिनव कुमार - Other Projects

3D Shape Recognition using Multi-view CNN

- report
- [Spring, 2016]
- Guide : Prof Ganesh Ramakrishnan, Department of CSE, IIT Bombay
- Team : Nikunj Patel, Saurabh Kumar, Abhinav Kumar, Chandra Mohan Sharma and Hitesh Tulsiani
- Trained two consecutive CNNs for recognizing 3D shapes from 12 independently rendered perspectives.
- Achieved 88.1% accuracy on ModelNet40 dataset with a simpler network architecture.
- Rated as the best project in the course CS725: Foundations of Machine Learning.

Color Constrained Delaunay Triangulation

- report
- [Spring, 2016]
- Guide : Prof Siddhartha Chaudhuri, Department of CSE, IIT Bombay
- Team : Abhinav Kumar, Tushar Kataria, Nikunj Patel and Saurabh Kumar
- Proposed fusing color information with spatial locations in Delaunay triangulation for point cloud data.
- Achieved a reduction of 50% in the number of triangles with different colors compared to vanilla Delaunay.

Total Variance Image Denoising using the Split Bregman Method

- report
- [Autumn, 2015]
- Guide : Prof Navin Khaneja, Department of EE, IIT Bombay
- Implemented Anisotropic Image Denoising Algorithm for images corrupted by Gaussian Noise.
- Tweaked the Split Bregman iterations to improve the Mean Squared Error.

Transductive SVM using CCCP

- report
- [Autumn, 2015]
- Guide : Prof Ganesh Ramakrishnan, Department of CSE, IIT Bombay
- Team : Abhinav Kumar, Chandra Mohan Sharma and Saurabh Kumar
- Improved the error rate of the SVM in semi-supervlised learning setting using the Concave-Convex Procedure (CCCP).

Retinal Vessel Segmentation using Wavelets and Supervised Classification

- report
- [Spring, 2015]
- Guide : Prof Suyash Awate, Department of CSE, IIT Bombay
- Team : Abhinav Kumar and Tushar Kataria
- Carried out edge detection of fundus images using Morlet Wavelet.
- Passed histogram of edges to SVM to classify a diseased and a normal eye with an accuracy of ~75%.

MAP Image Denoising using Markov Random Field prior

- report
- [Spring, 2014]
- Guide : Prof Suyash Awate, Department of CSE, IIT Bombay
- Carried out MAP Denoising with Quadratic, Huber and Discontinuity Adaptive MRF Prior using a 4-neighborhood.

Unsharp Masking of Noisy Images using Non-Linear Techniques

- report
- [Autumn, 2014]
- Guide : Prof Shabbir Merchant, Department of EE, IIT Bombay
- Team : Abhinav Kumar, Durgesh Singh and MPS Hemanth Kumar
- Carried out Cubic unsharp masking on noisy images that performed better than linear unsharp masking.

People Counting Solution for Efficient Evacuation of Large Buildings

- paper / patent (PERCOM Workshops 2015)
- [July, 2013 - Apr, 2014]
- Mentor : Avik Ghose, Innovation Labs, Tata Consultancy Services Limited
- Estimated smartphone user’s direction from the noisy raw accelerometer values in real time without calibrating the smartphone beforehand.
- Determined the phone placement by extracting novel features in the accelerometer data and passing them on to SVMs.
- Achieved an accuracy of 99.2% and 94.3% for phones with and without gravity sensor respectively.
- Placement dependent user direction was used to record entry and exit of people at the gates in an Android app.

Image Registration System on ARM Based Embedded Platform

- report
- [July, 2012 - May, 2013]
- Guide : Dr Kailash Chandra Ray, Department of EE, IIT Patna
- Estimated the affine transformation of two images using High Pass Emphasis Filtering, Log Polar Transform and Cross Power Spectrum Technique.

Cell Images Classification and Software Implementation

- report / demo
- [May - July, 2012]
- Guide : Prof Badrinath Roysam, Department of ECE, University of Houston
- Team : Abhinav Kumar and Vidit
- Used hue channel histogram to classify histopathology cell images outperforming gist features.
- Developed Qt/C++ based GUI for the Farsight toolkit for automated grouping and visualization.

FPGA Implementation of 1D Discrete Mellin Transform Processor

- report
- [Spring, 2012]
- Guide : Prof Kailash Chandra Ray, Department of EE, IIT Patna
- Team : Abhinav Kumar and GSS Praveen Chand
- Implemented 1D Discrete Mellin Transform Processor for eight samples on Xilinx Spartan 3E FPGA in Verilog HDL.
- Implemented sequential pipelined architecture in the processor.

PLC Water Level Controller

- report
- [Autumn, 2011]
- Guide : Prof Shovan Bhaumik, Department of EE, IIT Patna
- Team : Abhinav Kumar, Ajay AL and Subhash S.
- Made a water level controller using PLC Programming.

Technology Enabled Assessment Environment

- pdf / report / project page (IASTED 2011)
- [May - June, 2011]
- Guide : Prof Deepak B Phatak and Avinash Awate, Department of CSE, IIT Bombay
- Team : Surabhi Gupta, Danda Anusha, Alefiya Lightwala, Abhinav Kumar, Mayank Singh, Rohit Kondekar and Nishi Kumari
- Open Source Project that enabled taking Multiple Choice Tests on Java based Bluetooth enabled mobiles and get the score instantaneously upon submission.
- Handled the Assessment Module. Languages used were Java and HTML.
- Implemented a generic tree with objects as its nodes. Created parent node results by traversing the tree till its root.
- Informed students of their subject-wise ranks and other performance metrics.

Car Parking System

- report unavailable
- [Autumn, 2010]
- Guide : Prof Ranjan Kumar Behera, Department of EE, IIT Patna
- Team : Abhinav Kumar and Abhay Kumar Priyadarshi
- Designed a digital system which monitors available space in a parking garage for eight cars and displays least vacant parking place, the numbers of cars parked and whether the garage is full.
- Used VEM technique to simplify the truth table and implemented the logic using basic gates ICs

Google Sites

Report abuse