Software Engineer @Simply Business || Ex- SDE @Amazon UK || Ex-SWE Intern @ Microsoft India, Sourcewiz || Ex-NUS Intern 

github

Ritik Agarwal  • Peerlist

Implemented a full Pipeline Model which First detects if The Image is of a person or Dog.If its a Dog,the model predicts what is its breed from 133  classes of breed.Implemented the Model in PyTorch Framework using Both ie from Scratch as well as using Transfer Learning.For transfer Learning the pre-trained model taken was ResNet-50.The testing accuracy Got from Transfer Learning model was 80 %.

Dog Breed Classifier Using Deep Learning

Implemented Different Pretrained Models like UNET with Different Pretrained Encoders and Ensembled them to give final prediction

Vegetation Region Extraction using Semantic Segmentation

Built a Neural Network from scratch Using the NumPy Toolbox only.The goal of this project was to learn the absolute basics of what goes on behind some of the common operations achieved using common Deep Learning Frameworks like PyTorch,to gain a deeper understanding.

Predicting Bike Sharing Data

Implemented CNN model(tiny VGG) with 5 Convolutional Layers to classify Images into 5 emotion categories-
Angry,Happy,Surprised,Sad,Neutral.Got a training accuracy of around 70 percent and Validation accuracy of around 60 percent.Implemented using OpenCV library ,to get real time lmage  of Face from Webcam and Give the output as which emotion class it belongs to.