Hand Recognition program to play the dinosaur game on Google Chrome. I generated multiple images of my hand in open and closed position using OpenCV. I extracted the brown colour of the hand from camera feed by creating a HSV mask in the range of brown colour and combining the same with the original image. I then added code to train a simple Convolutional Neural Network using Keras on these generated images to identify the hand position. I wrote a different program to capture camera frames and run it through the CNN to identify the position and accordingly make the dinosaur jump.

Link: github.com/ckarthik14/Hand-Gesture-Recognition

Automated the billing process in supermarkets using RFID technology integrated into a smart trolley which is connected to an AWS Relational Database System. Built a website using the Django framework connected to the AWS RDS instance to display item and billing data in real-time to the users. Simplified the process of adding items and removing items from the cloud and billing on exiting the store using a Raspberry Pi micro-controller for integrating the various components such as an LCD screen, RFID Reader, and RFID tags.

Link: ieeexplore.ieee.org/document/8960144

Trained and Compared accuracies of standard Convolutional Neural Networks like VGG16, ResNet50, and Xception on the Google Landmark Recognition Challenge dataset. Performed transfer learning using pre-trained "ImageNet" weights to exploit existing knowledge regarding edge detection, feature extraction, etc. to achieve a high accuracy in all three networks.

Link: github.com/ckarthik14/Landmark-Recognition

Trained a simple sequential neural network resembling an XOR gate initialized with random weights. Achieved a high accuracy for all input combinations of a standard XOR function.

Link: github.com/ckarthik14/XOR-Problem