Rushikesh Tade

This is firefox plugion which downloads all the files from current HTTP directory. It arranges downloaded files in an order which is similar to server. All the files are stored at Downloads/HTTP Directory/*

This project was aimed towards creating an image augmentation platform which will help other data scientists to solve problems class imbalance or of less samples. This platform could reproduce each image into 218 different images using various image processing transforms.Result will be the zip file containing images that are available to download to user.

Digital twin of ball mill was created in order to simulate effects of various factors in its design as well as operation. Design aspects considered were Mill Diameter, Mill Length,Vertical Drop etc. and Operating aspect were Mill RPM, Feed Size etc.

Call context was analyzed with one of the car manufactures where incoming calls were recorded and then converted into text using Watson STT. This calls then feed to call context analyzed which was showing the context of call interms of car parts & its urgency. Results of various calls were displayed on a dashboard in order to give this results as a feedback for R&D and Quality team.

Firefox addon to download high resolution images from google image search.

Elevator feed was analyzed to detect unattended object , number of people in elevator and people distribution according to ages so that ads can be targeted according to audience.

Lobby feed was analyzed in order to detect people in lobby so that lift can be called even before used pressing button which will reduce waiting time in lobby area.

This project was aimed at integrating Watson conversations with Skype for Business on premises installation. Microsoft LinQ Server Libraries were used in order to automate Skype chats.

ARIO is an image search application made with Unity3D. It has features such as attaching data with the image results.
Data can include various types such as video clips ,images or even 3D objects.

Human Detection in images is performed in this experiment. HOG and BO features are used for the experimentation. First logic was developed in Matlab on it was ported into C++ with OpenCV. We have used GPU functionality of OpenCV in this project.

Languages/Technologies/Libraries Used::MATLAB ,C++ with OpenCV

In this project we have detected the theif using RGB-D Videos recorded by Kinect. Actions were performed by 3 persons and 4 videos each in each category. Results has given accuracy of around 94 %.

Languages/Technologies/Libraries Used: Matlab, Kinect, nuiCapture,kNN,PCA.

In this project, an accelerated JPEG (Joint Photographic Experts Group) resizer is to be efficiently implemented on the GPU (Graphic Processing Unit) using the CUDA (Computer Unified Device Architecture) technology and it is capable for high definition images decoding. The CUDA technology can assist the GPU to work for the CPU for large computation. In this project, the DCT (Discrete Cosine Transform) and IDCT (Inverse Discrete Cosine Transform) model which consumes about 75% the total time in… Show more

In this project, an accelerated JPEG (Joint Photographic Experts Group) resizer is to be efficiently implemented on the GPU (Graphic Processing Unit) using the CUDA (Computer Unified Device Architecture) technology and it is capable for high definition images decoding. The CUDA technology can assist the GPU to work for the CPU for large computation. In this project, the DCT (Discrete Cosine Transform) and IDCT (Inverse Discrete Cosine Transform) model which consumes about 75% the total time in the JPEG decoder is worked in the GPU by the CUDA, and other models of the JPEG are to be designed in the CPU. At the same time, using the asynchronous parallel execution between the CPU and the GPU to improve the JPEG resizing acceleration rate. In the project, the JPEG resizer based on the CUDA performs resizing of 4K HD images of size 3240 x 2160 pixels images.

Languages: CUDA C
Show less

In this project we have implemented few anti-aliasing algorithms into MATLAB such SMAA and MSAA.

Languages/Technologies/Libraries Used:Matlab

In this project text is detected in the Videos and Images. Using morphological operations and combination of different filters on rotated images we have detected the text.

Languages/Technologies/Libraries Used: Matlab

An enhanced fuzzy min–max (EFMM) network is implemented in this project. The aim is to overcome a number of limitations of the original fuzzy min–max (FMM) network and improve its classification performance. In this project EFMMNN is implemented using MATLAB and later it is converted into JAVA.The Testing is done on FisherIRIS dataset giving results upto 96% .

Languages/Technologies/Libraries Used: Matlab , Java

OligoIs is the instance selection algorithm used for the class imbalanced dataset problem. We have performed experiments on around 9 datasets in UCI repository.

Results has shown that it gives a balanced dataset which gives the recognition ratio around 90 % if tested using kNN.

Languages/Technologies/Libraries Used:Matlab,Java

Generating models from large data sets—and determining which subsets of data to mine—is becoming increasingly automated. However, choosing what data to collect in the first place requires human intuition or experience, usually supplied by a domain expert. In this project a new approach to machine science which demonstrates for the first time that non-domain experts can collectively formulate features and provide values for those features such that they are predictive of some behavioral outcome… Show more

Generating models from large data sets—and determining which subsets of data to mine—is becoming increasingly automated. However, choosing what data to collect in the first place requires human intuition or experience, usually supplied by a domain expert. In this project a new approach to machine science which demonstrates for the first time that non-domain experts can collectively formulate features and provide values for those features such that they are predictive of some behavioral outcome of interest. This was accomplished by building a Web platform in which human groups interact to both respond to questions likely to help predict a behavioral outcome and pose new questions to their peers. This results in a dynamically growing online survey, but the result of this cooperative behavior also leads to models that can predict the user’s outcomes based on their responses to the user-generated survey questions.For prediction we have used Linear Regression and Naive Bayes model.

Languages/Technologies/Libraries Used: ASP RAZOR V3 C# Show less

In this project Public sentiment analysis on twitter is done. Tweets were searched for the keywords like Modi, Apple,Congress.Results were showing the sentiments category towards the topic.For the Tweets retrival Twitter4j library was used and for the sentiment analysis SentiStrength was used.

Languages/Technologies/Libraries Used: JAVA , SentiStrength, Twitter4j

Gait Recognition i.e. recognition of persons by his walking style is done by using TUMGAIT Dataset. We have archived Recognition ratio of around 65 % in this project.

Languages/Technologies/Libraries Used:MATLAB

We have designed the website for the virtual Body. Users were supposed to put the annotation on the body.

Languages/Technologies/Libraries Used:PHP, Javascript, JQuery, PHPmyAdmin, Parse, Annotorious , Microsoft Azure.

We have created a dataset for the 3D action recognition system using Microsoft Kinect.Dataset was recorded for 10 persons and 11 Videos/persons using nuiCapture. Then with the combination of SVM, NN, KNN, Naïve Bayes classifier we have achieved an Accuracy up to 96 %.

Languages/Technologies/Libraries Used: MATLAB, Kinect, nuiCapture, Neural Network, Support Vector Machine, k-Nearest Neighbor, Naive Bayes.

In this project new image retrieval algorithm that establishes the active interaction between the user and the system. Extraction of the features of an image used in this proposed algorithm is controlled by the need of the user. The various classifiers based on Artificial Neural Network are the special attractive features of this algorithm. These classifiers helps the system to maximize the satisfaction of the Users. The Genetic Algorithm as an Evolutionary Computational technique is deployed… Show more

In this project new image retrieval algorithm that establishes the active interaction between the user and the system. Extraction of the features of an image used in this proposed algorithm is controlled by the need of the user. The various classifiers based on Artificial Neural Network are the special attractive features of this algorithm. These classifiers helps the system to maximize the satisfaction of the Users. The Genetic Algorithm as an Evolutionary Computational technique is deployed here for optimizing the retrieval process in the form of Relevance Feed Back in the system. The algorithm is tested on the huge collection of diversified Corel Image Data base. The Image classification and indexing is completely characterized by the integration of the low level features such as multi-resolution color, Local Binary Pattern (LBP), Histogram Oriented Gradient (HOG), Scale Invariant Feature Transform (SIFT), Edge Histogram and texture features. The experimental result shows that the performance of this Image Retrieval system is qualitatively improved than the existing similar algorithms.

Languages/Technologies/Libraries Used :MATLAB, Vlfeat ,SIFT, LBP,Genetic Algorithms , HOG ,EDH, Neural Network ,SVM , kNN. Show less

A finger print recognition system was implemented in MATLAB for detection of the persons, Minutia features were used from Fingerprints and Back Propagation Neural Network was used for the classification.

Languages/Technologies/Libraries Used: MATLAB

In this project we have performed the clustering of the comments towards political party supporters in election period. Comments were extracted from times of India Website.Variational inference DPMM based clustering method was used for the clustering of comments as positive,negative or neutral.

Languages/Technologies/Libraries Used : MATLAB, JAVA.

It was a simple security system which used to greet the new guests and then it was recording their voice and capture a photo and send a mail to personal email Id. We had also attached the USB GSM modem HUEWEI E150 for sending the SMS of guest arrival.
Languages used: Perl, Bash Scripting.

In this project I have created a karaoke player using raspberry Pi. The audio card called as Wolfson Pi is used for the clearer and better output.

Languages/Libraries/Technologies Used :Python , Pygame , Raspberry Pi, Wolfson Pi, Debian Linux

Developed the emotion recognition system using Matlab. Multiple waveforms were taken for the testing into four types of emotions. Results are achieved via Polynomial Classification.

Languages/Technologies/Libraries Used: MATLAB

Segmented of the Knee in the MRI images using Aactive Shape Models and Active Appearance models.

Languages/Technologies/Libraries Used: Used:MATLAB

Speaker Recognition was performed using MATLAB, MFCC with polynomial classifier was used for the experiment.Recognition ratio was around 90 % for this experiment.

Languages/Technologies/Libraries Used: Matlab , Polynomial Classifier , Neural Network.

In this project the ontology model for the Marathi language was developed for answering the questions.
Paragraph about Shivaji Maharaj was selected for the answering.

Languages/Technologies/Libraries Used: Matlab

Speech recognition experiment was performed in this project first MATLAB code was developed in which MFCC and KNN were used for recognition. Later on this project was converted into the Embedded C for TMS320 Platform in which speech recognition was done.

Languages/Technologies/Libraries Used: Used:MATLAB, Embedded C, Code composer Studio.

Conducted a four days’ workshop on Advanced Micro-controllers at Veerayatan Institute of Engineering as an instructor, All the Examples for 8085 and ARM7 and DSP TMS320 series were developed by us.

Languages/Technologies/Libraries Used :Embedded C,Code Composer Studio , Keil uVision , Microcontroller 8051, DSP TMS320 6713, ARM7 LPC2148

We have developed an action recognition system using KNN classifier and LDA , PCA combination. Final Recognition ratio was 76% .

Languages/Technologies/Libraries Used : MATLAB, kNN , PCA ,LDA , Video Processing

To reduce the piracy along with providing the user reliable and safer way to access the application. As an embedded platform raspberry pi with Wheezy Linux was used packages such as PHP5 were used for building the cloud environment. At last the device was connected to the network, it was able to serve the purpose of it as a cloud based compiler for C/C++ and Python.

Languages Used:PHP,JavaScript,Linux Shell Script,C,C+,Python.

Simple unidirectional visitor counter was created using Time IC, Interface to PC was given by using mic jack. In MATLAB a pulses were counted using the incoming audio signal.

To convert the single phase supply 220v into the three phase supply 120 volts each. Darlington Pair Transistors were used for the conversion of the single phase to three phase.

This Firefox plugin lets you view/render python notebooks without running notebook server. Once installed users simply needs drag and drop .ipynb file into firefox and it will be rendered.

This project consist of Natural Language processing on various equipment manuals. Based on manuals a knowledge graph is created which can be used to answer question regarding issues in equipments such as what is cause of issue, how to solve this issue and description of issue.

This project was build in partnership with ABB systems where their robot YuMI was using computer vision to detect slot status on PCBs such as empty slot, incorrectly filled slot or occupied slot.
It was also using IBM technologies for Interacting with users for training of Assembly procedure.

This project was featured in one of the ABB's Expo in Switzerland for international audience.

Answer Anything is a web scrapper application which is used a backup for Watson Chatbot and Watson Discovery Services. Whenever user puts a question which is not available in Watson conversation dialogs or in Watson discovery service then that question can be redirected to this application. A GET API was provided in order to easily integrate this into existing project. Answer Anything will simply search on Google and parse the answer to user for questions which could be almost anything.

Digital Factory AR was an android augmented reality application developed with 3D engine Unity and AR Framework Qualcomm Vuforia. Idea was to put VuMark (similar to QR codes but with Graphics) on different pipes and different part of factory. Whenever Plant manager or maintenance person will come he can simple glance the status of each part or pipe simply by pointing camera towards that part.

In this project stones which are on converyour belt are processed using computer vision . Based on feed analysis stone quality and quantity is predicted which is then displayed on the dashboard.

It is prototype development of Virtual Camera in 3D scenes. It is suppose to be helpfull to people working in animation industry where real like camera is required to get shot. Result of this product was to a video which required minimal post processing time.

Developement of basic AR headset. It was based on SLAM engine which was helping end product to detect collisions. We also able to reconstruct the environment on PC platform in realtime.