Name: Megh Shukla
Type: User
Company: Ecole Polytechnique Federale De Lausanne
Bio: Computer Vision Ph.D. Student @vita-epfl |
ex-Research Engineer, MBUX Interior Assist, Mercedes-Benz
Twitter: MeghShukla
Location: Lausanne, Switzerland
Blog: https://meghshukla.github.io
Megh Shukla's Projects
Author implementations for VL4Pose, EGL++ and LearningLoss++, and "unofficial" implementations of various active learning algorithms for human pose estimation!
Programming of Firebird V Hardware development platform with Atmega 2560 Microcontroller.
GUI implementation with CUDA kernels and Numba to facilitate parallel execution of Maximum Likelihood and Relaxation Labelling algorithms in Python 3
Processing Data with class imbalances, missing values for Machine Learning applications. Use of cross-validation and it's limitations
Neural Network coded from scratch, only library used is numpy; implemented as a part of my BE project.
Jupyter notebook visualization of 2D single person human pose estimation.
Dimensionality Reduction and visualization technique that compensates for the curse of dimensionality (ICASSP '20)
A beautiful portfolio Jekyll theme that works with GitHub Pages.
Loading and Visualizing MPII Human Pose dataset in Python
Principal Component as a tool for Dimensionality Reduction using Hyperspectral images
Code repository for "TIC-TAC: How To Learn And Evaluate Your Covariance". We derive the TIC: Taylor Induced Covariance and show that incorporating the gradient and curvature results in better covariance predictions. We also propose TAC: Task Agnostic Correlations, a new metric to quantitatively evaluate the covariance.