Harekrissna's repositories
Buried-threat-detection-using-AI-on-GPR-data
We, Achin and Harekrissna worked as a team to complete the project given to us on Buried threat detection using ground penetrating radar. We applied Deep Learning techniques specifically CNN and transfer learning along with image processing techniques like color thresholding, augmentation and masking to identify the threats hidden underground by analysing the radar data. We implemented the techniques given in the research paper (Some Good Practices for Applying Convolutional Neural Networks to Buried Threat Detection in Ground Penetrating Radar, by Daniël Reichman, Leslie M. Collins, Jordan M)
EmoTV
As the name suggests, EmoTv is capable of interacting with a person and determine his current mood. Running on python coded lines, it can efficiently involve in a conversation with a subject. Our TV features playing songs alongside the emotion recognition part. As an added bonus, it shows up a cartoonized image of yours with your name written on bottom right of screen and displays your emotion. We have also added image magick gui alongside the cartoonized image for further editing of the image. The entire project is completely based on python scripts, displaying the vastness of the language. Our project covers topics ranging from machine learning using Tensorflow to Speech recognition and conversion. Our EmoTV Project is a step towards making robots understand human emotion, thus lessening the gap of human machine interaction. Team name: VCare Team members name: 1) Rathod Harekrissna Upendra (Leader) 2) Parth Shettiwar 3) Adhishree Apte 4) Ayush Raj We hope to further improve upon this project by involving gesture recognition and more conversation. Also, we request access to some good dataset of Indian faces with varied emotions regarding which, help will be much appreciated.
Hyperspectral-Image-Clasification
Hyperspectral Image Classification through Softmax followed by KNN
Optimal-Network-Allocation
In the modern world, using cellular communication has become a necessity and with growing demand for networkconsumption, the number of base-stations providing the facility has also increased considerably. With this huge number ofusers and base-stations providing the service, choosing an optimal allocation of all users to base stations has no longerremained a simple problem. There are various research work done along similar problems already like one approach usedby Prof. Prasanna Chaporkar in his paper on Joint Cell Zooming.Utilizing this opportunity, we decided to try and solve this problem as an Research and Development project un-der Prof. Prasanna Chaporkar. We start with defining the problem statement, then we analyse the complexity of theproblem - the problem turns out to be NP-Hard. In order to achieve globally optimum solution, we can not do muchbetter than brute-force if we want a deterministic algorithm to find the solution. We had initially tried to solve it using agreedy approach, but we could easily come up with a small sized counter example where in the greedy approach producesa result that is a local optimum but not a global optimum. The example is presented in the simulation section, and wealso show that our other approach settles at the global optimum. Hence, motivated from approach used by Prof. PrasannaChaporkar in his paper on Joint Cell Zooming, we also decided to formulate our problem as a Potential Game and solveusing a probabilistic approach, that is Spatial Adaptive Play (SAP). This approach adds a certain degree of randomnessto regular BRD algorithm so that it doesn’t get stuck in local optima.
Faster-RCNN
This is an implementation of Faster RCNN to detect underground threats in GPR data as a continuation to my internship at TATA Power SED
Flappy-Bird-AI
Used Genetic Algorithm of Natural Selection to build an agent capable of playing Flappy Bird game and obtain a score of over 500 pipes at 60fps
Medical-Image-Segmentation-using-UNET
A short analysis on strategizing the hyper-parameter tuning on the task of Image Segmentation using UNET on Medical image data
Reinforcement-learning
Implementation of various traditional Reinforcement learning algorithms
Sedrica-path_planning
Part of a mega project Self Driving Car (SeDriCa) by UMIC-IITB as a part of competetion Mahindra RISE driverless car challange. This repository contains the path planning module in which I contributed
CBC-Encryption-with-AES-32-security
This is a course project. We have implemented a fully functioning and deployable algorithm of CBC mode with AES-32 encryption and decryption purely in python using no library other than numpy.
hello-world
My first repository
Path-planning-data
Just a repository of data corresponding to path planning subsection which we use for training. Contains bag files and python codes to use them