Derek Gloudemans's Projects
Gradient descent-based optimization of an input truss using matrix-based stiffness method for structural analysis.
Trials of formulations for 3D object detectors, as well as a 3D object dataset
A Blokus game interface in which a variety of heuristics, search strategies and learning agents are implemented. Deep RL state estimation will be added when I get some free time!
A few classifiers are implemented for the challenging ISIC 2018 skin lesion classification challenge.
Little snippets of code to remind myself how things work
Implements 2D and 3D perspective transform between image and cartesian coordinate plane, and curvilinear to cartesian coordinate conversion. Useful for coordinate systems defined nominally relative to a single non-linear axis (e.g. roadway stations and mile markers)
Utilities for object detection, object tracking, trajectory conversion from image space to world coordinates, and track plotting.
Homography, plotting and detection for dashcam video footage and other ego-vehicle data
Pytorch implementation of deep deterministic policy gradients for reinforcement learning over continuous action spaces
Repository for most of the code from my YouTube channel
A simple and straightforward implementation of a decision tree classifier as described by Leo Breiman in Classification and Regression Trees.
Portfolio website of major research and personal projects
Performs stream processing (nominally object detection) across multiple worker processes with decentralized load balancing, decentralized database with eventual consistency, and basic fault monitoring.
A C++ DNA_strand object is implemented by dynamically allocated array and by linked list, and functions are provided for searching strand for a sequence, appending to it, cleaving target sequences from it, etc. Comprehensive testing scripts also implemented.
Large-scale sparse trajectory annotation with all the bells and whistles, with goal of labling one vehicle in 20 cameras in 1 minute. Pulls a majority of code from other annotation tools I've built with a few bells and whistles optimizing for this task specifically
An implementation of basic hierarchical tree (dendrogram) construction with a few modifications and enhancements, applied to a fun dataset of beers.
Code, data, and video examples from I-24 MOTION feasibility test.
Utility functions for loading images from the KITTI dataset, plotting 2D and 3D bounding box and evaluating outputs according to the KITTI dataset metrics
Simple implementation of lambda means clustering
Fast Vehicle Turning-Movement Counting using Localization-based Tracking
LBT-count without torch.multiprocessing to enable sharing of CUDA memory between processes. Instead, images are transferred to CUDA on main process
Very fast tracking and detection using a CNN-based localizer and Kalman filter.