GANS for Data Augmentation

- Developed and validated performance of a Deep Convolutional Generative Adversarial Network for Data Augmentation using Pytorch.
- Generated over 2000 Artificial training data from the Standford cars dataset

Team STORKS - Apartment Package Delivery with UAV

- Currently working on a drone to deliver packages to apartment balconies autonomously.
- Implemented a Real Time Obstacle detection system based on YOLO, trained using self annotated custom dataset.

3D Reconstruction and Bundle Adjustment

- Performed eight-point and seven-point algorithms to estimate Fundamental Matrix.
- Computed epipolar correspondences to obtain 3D Reconstruction of a given scene.
- Optimized reconstruction leveraging RANSAC and Bundle Adjustment.

Mahindra Rise Driverless Car Challenge

- Worked on modelling and simulation of Ego vehicle utilizing ROS, Gazebo and Rviz to test sensors, communication and making system fault tolerant.

Optical Character Recognition

- Developed OCR pipeline to recognize handwritten alphabets and spatial information such as spaces and new lines. - Implemented a CNN in Pytorch and obtained an accuracy of >90% using OCR on test images.