Sarosij Bose's Projects
An late fusion architecture between I3D and Xception for Human Activity Recognition
This is the pytorch implementation of some representative action recognition approaches including I3D, S3D, TSN and TAM.
Algorithm for 2D Bodytracking
Here a deep learning model is built to predict whether an image is that of a dog or a cat from it's image using the famous MNIST dataset.85.1% accuracy is achieved.
This Repository detects and classifies Diseases in leafs as well as uses the FiftyOne tool to visualize various data parameters.
Introduction to Parallel Programming class code
Data Free adversarial defence at test time
IoT course capstone project
This project aims to predict to which species of Iris family does a flower belong to among three species:Iris setosa, Iris virginica and Iris versicolor.A ML model is created which uses the Iris flower dataset containing 50 samples of each flower or a total of 150 samples. The model uses a SVM for classification and also runs a Gridsearch for finding the optimal parameters.98% accuracy is achieved.
Overview of Lipschitz regularization of Neural Networks (FCNNs and CNNs)
Here a ML model is built to predict whether a borrower will be able to payback a loan or not using real world data from LendingClub.com from 2007-2010.
This repository contains some basic set of questions and answers for ML interviews
ResCNN: An alternate CNN implementation
This is a NLP/ML model which aims to predict what rating will a customer give a business listed on Yelp in the future.
This repository contains the [PyTorch] implementation of the paper: "Towards Frequency-Based Explanation for Robust CNN"
My personal Information
PySlowFast: video understanding codebase from FAIR for reproducing state-of-the-art video models.
A unified framework for KD applied to Sport Analytics
Generates the product of loose upper bound of each layer for a trained neural network
This repository contains an implementation of Traffic Sign Classification using the OpenVINO toolkit Inference Engine.