Some of my academic and weekend projects. Please visit my GitHub for more.

• Federated Learning - MNIST / CIFAR-10 (Jul 2021) - Basic starter code using Keras / TensorFlow for simulating Federated Learning on IID and non-IID MNIST datasets. This Kaggle notebook includes code for creating non-IID dataset using the Sort-and-Partition method and for random client selection for every communication round.

• Brain Tumor Segmentation from MRI using PSPNet (Apr 2021) - Motivation was to show the efficacy of PSPNet using Keras for the segmentation of brain tumor from MRI data. IoU of 0.552 and Dice coefficient of 0.66 were achieved.

• Car Price Prediction ML WebApp (Mar 2021) - Built a machine learning web application for predicting price of used cars using Random Forest Regressor model. Implemented using Python, Flask, HTML, and Bootstrap and hosted on Heroku cloud.

• Cell Nuclei Image Segmentation using U-Net - Cell nuclei images and their corresponding ground truth masks were used to train a U-Net like model and an acccuracy of 95.17% was achieved.

• Sentiment Analysis of Amazon Book Reviews - A comparison of different ML models for classification of Amazon book reviews into positive and negative classes. Linear SVM performed the best. Further, GridSearchCV was used for hyperparameter tuning of Linear SVM.

• Multi-Label Classiffication of American Sign Language using Attention based Transfer Learning on VGG16 - Hand sign images for 29 classes were used to train a VGG16 based Attention model and an acccuracy of 84.5% was achieved.

• Quadrone, Cognizance (Mar. 2018) - Participated in the national technological fest of IIT, Roorkee with a team of four members. The flight controller of a remote-controlled quadcopter was implemented using Arduino UNO, MPU6050 and other off-the-shelf components. Flysky CT6B 2.4GHz 6CH Transmitter was used as the radio controller. Learnt about quadcopters including PID controllers, radio communication and embedded processing.