Lattice Boltzmann CFD with
Machine Learning (Python)
Uses Lattice Boltzmann method to create a fluid simulation of objects. Used simulation to create dataset, which was then used to train a nueral network to predict air flow. Training was largly unsuccessful due to lack of compute.
Android Application that uses Tensor Flow's Single-pose Thunder convolutional neural network to estimate joint positions of human body. Then uses pose classification model to identify phase of running motion (floating or touchdown).
Uses Dijkstra's algorithm to find an optimal path through a maze, and turns that path into a set of commands. A PD controller with tuned parameters is used to control the robots movements.
Performs a stereo calibration to determine intrinsic and extrinsic parameters for stereo vision setup. Uses parameters to determine 3D positions of objects of interest.
Building two traffic lights designed for a T intersection. Traffic light behavior is determined by a state machine running on a Nucleo board with inputs from two ultrasonic sensors.
Exploring two different ways to blend two images. Laplacian blending involves using a mask to blend high and low-frequency information across two pictures as seen above. Hybrid images take the high-frequency information from one photo and the low-frequency information from another and merge them into one.
Using temporal averaging and morphological operators, a connected components analysis is used to distinguish between separate objects and track their locations over time.
Passes a smooth curve through a set of points using piecewise third-order polynomials. Constructed linear system and implemented Gauss-Seidel Method for solving it