Transfer Learning with Augmented Vocabulary for Tweet Classification
Vineet Kumar, Karthikeya Racharla, and Dr. Debapriypo Majumdar
Sixth IEEE International Conference on Multimedia Big Data 2020 (best paper award)
paper / slides / code / video

In this work we describe our approach for domain specific (#MeToo movement) tweets classification with severe class imbalance into five linguistic aspects. Our approach ranked first in IEEE BigMM Grand Challenge.

NTIRE Real Image Denoising: Dataset, Methods and Results
Abdelrahman Abdelhamed, Mahmoud Afifi, Radu Timofte, Rajat Gupta, Vineet Kumar
IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops 2020
paper

This work describes our approach to NTIRE 2020 challenge on real image denoising.

Predominant Musical Instrument Classification based on Spectral Features
Karthikeya Racharla, Vineet Kumar, Chaudhari Bhushan Jayant, Ankit Khairkar, and Paturu Harish
7th International Conference on Signal Processing and Integrated Networks (SPIN) 2020
paper / slides / code

This work aims to examine musical instrument classification. We used musical clips recorded from various sources and built supervised and unsupervised models using spectral features.

Distributed Approach for Batchwise GCD Computation of RSA Moduli
Vineet Kumar, Aneek Roy, Sourya Sengupta, Dr. Sourav Sengupta
13th International Conf. Information Systems Security 2017 (acceptance rate: 30%)
paper / slides / code

In this work we propose a distributed method for common factor attack to RSA Moduli

Acoustic Features to Predict Topic Change in Instructional Video
Vineet Kumar, Sushant Gupta, Sonal Patil, Dr. Om Deshmukh
Presented at XRCI OpenHouse Bangalore | July, 2016 | [Slides] | [abstract]

Instructional Videos have given an open challenge to all current methods of education system. They have high potential of acceptability among all kind of learners. Instructional Videos have become a label for many recent course initiatives from higher education institution. We believe that Instructional Videos are going to be next generation textbooks. The Goal of this Project is to automatically identify the locations where the topic has changed in an instructional video. We specifically want to explore techniques that use acoustic (i.e. spoken) features.

Material Understanding in 3D Models using Kinect
Vineet Kumar, Chetan N.

Material Understanding in 3D Models using Kinect: Captured stereo RGBD images using Microsoft Kinect v2. Used InfiniTAM Reconstruction framework to build 3D Reconstruction model using captured color & depth images. Integrated a CNN based semantic segmentation library with InfiniTAM Model to understand the material in the reconstructed 3D model.

NMF based method for Aspect Extraction from Product Reviews
Vineet Kumar, BTP

With advancements in digital communication & security online shopping has garnered a significant pace since last decade. Most shopping websites allows buyers to share their feedback (reviews) about their purchased product. Since there are lots of reviews & they are mostly unstructured text – it is difficult to employ any data mining techniques to get interesting (non-trivial, implicit, previously unknown and potentially useful) information. It is also difficult to build an opinion mining application to compare various product models to make a purchase decision. These comparisons should be based on the features or aspects of the products. In this work, we explored Non negative Matrix Factorization based techniques to extract aspects from the raw reviews.

SIC Assembler Emulator
Hari Bhushan, Sushant Gupta, Vineet Kumar, Shubham Kr Ranu
Course Project, Fall 2016

Programmed and debugged the fully functional command line and GUI based simulator for SIC Assembler. The Simplified Instructional Computer (also abbreviated SIC) is a hypothetical computer system that is used to learn systems programming.