🔗 GitHub
Description
Guide: Prof. Dr. Pratyush Kumar (Assistant Professor, Dept. of Computer Science, IIT Madras)
During my internship, I worked under the guidance of Prof. Pratyush Kumar (Assistant Professor, Department of Computer Science, IIT Madras) where I implemented a Convolutional Neural Network for 6-DoF Global Pose Regression and Odometry Estimation from consecutive monocular images. The model estimates the camera pose from a sequence of monocular images from the camera. At each step, the model takes two consecutive frames as input and returns the global and relative pose between the two frames. It was built and trained from scratch in Tensorflow and it outperformed traditional feature-based visual localization algorithms, especially in texture-less regions. The neural network was later used by Prof. Pratyush for the localization of robots in GPS denied environments.
