Project Description:
This was my research project at ISRO. This project has direct linkage to the RISAT-1 mission of ISRO. The aim of this project was to find a method to quantitatively estimate the value of scalloping or banding noise present in an ScanSAR image. I had succesfully made an algorithm for estimating scalloping or banding noise present in an image. Moreover, I had also made a software for the same using python and tkinter.
During the project since I was not able to go to ISRO due to lockdown, I was not able to work on image present at ISRO. So, I had also made python program to simulate banding/Scalloping noise in an image using fast fourier transformation in the frequency domain.
Technologies used:
Python, OpenCV, NumPy, GDAL, Scipy, etc.
Project Description:
In this project our aim was to make a website which uses Bi-LSTM model at backend to predict fake news regarding corona virus spread on the internet.For, this project our team created data for training of Bi-LSTM.We also experimented with Machine Learning techniques like Naive Bayes and Support Vector Machine (SVM) and Deep Learning techniques like LSTM and Bi-LSTM for prediction .The best accuracy we received from Bi-LSTM model which was about 76%. For making website we used HTML , CSS , Bootstrap and Flask.
Technologies used:
Machine Learning (Naive Bayes, Support Vector Machine), Deep Learning (LSTM, Bi-LSTM), flask, HTML, CSS, Bootstrap, Python
Code See it workingProject Description:
In this project I created a huge dataset consisting of around 50,000 images of total 25 different classes:
The dataset was created by combining various datasets from kaggle. Then, I used transfer learning techniques for making classification model. I used VGG16 and got model with accuracy of 90.72%. I used Mobilenet V2 and got model with accuracy of 94.30%. I used InceptionV3 and got model with accuracy of 94.04%.
Technologies used:
Deep learning, Transfer learning, Tensorflow, Keras, Python
CodeProject Description:
In this project I created my own convolutional neural network using the concepts of InceptionNet and ResNet. Here, I used “Face Mask ~12K Images Dataset” from kaggle for training my convolutional neural network. I got accuracy of 97.78%. Then, I used two approaches to get faces from an input image. In first approach I used Multi-task Cascade Convolutional Neural Network (MTCNN) for extracting faces from image and then used my model to classify that person is wearing mask or not. In second approach I used Haar Cascade classifier for extracting faces from image and then used my model to classify that person is wearing mask or not. I also made python program to use my model on my laptop’s webcam for face mask detection.
Technologies used:
Deep learning, OpenCV, Tensorflow, Keras, Python
Code See it workingProject Description:
In this project our goal was to analyse and make predictions of temperature and humidity using Machine learning.In this project we used Raspberry pi and DHT11 sensor to gather the temperature and humidity data from surrounding and then we transfered the data to our computer using client server communication python program and then made predictions with linear regression machine learning algorithm and displayed the result on website (which we created using HTML, CSS and flask framework).
Technologies used:
Internet of things (IOT), Machine Learning (sklearn) , Flask , HTML , CSS , Bootstrap