Projects
The sections below highlight projects I've worked on over the past few years. This includes a pending publication and an award-winning presentation. Additionally, I included four other projects that I have made available to anyone. These are posted on my GitHub which covers topics on Security, Deep learning, and Data Visualization.
New Research | Pending Publication
Realtime ML-Assisted Data Collection using Drones
2022
This paper presents a system for realtime visual data collection with drones. The novel aspect lies in the system coordinating lightweight ML models and drone controls that help operators collect data efficiently and economically. ​
​
This image shows 5 examples of the system software detecting rip currents in realtime.
Benefits:
-
The collection of data can be done by drone operators who are not experts in the subject domain of the data.
-
The system is lightweight and customizable, allowing it to be modified and adopted immediately.
Demonstrated Applications:
Note: tested in an environment with no WiFi access and limited computational power availability.
-
A simple flight over the beach with an ML model trained to detect people. The associated mobile app is instructed to provide a realtime count.
-
Complex application with an ML model trained to detect two types of rip currents. Upon detection, the drone will hover above the rip current capturing a video.
Technical Details:
-
A convolutional neural network (CNN) detects the objects of interest via a serverless connection to a mobile device while the drone is in flight. This is achieved by integrating the lightweight CNN model with the drone's controller, which fits the constraints of limited mobile computing resources.
Award Winning Presentation
Machine Learning (ML) Supported Mobile Apps for Citizen Science
2022
Received award for Best of the Baskin School of Engineering following a presentation at the UCSC Graduate Research Symposium. This project is an extension of Realtime ML-Assisted Data Collection using Drones that seeks to make a Machine Learning template to assist researchers in developing new Citizen Science mobile and drone applications. Was a great experience communicating research to students, faculty, and judges.
Pictured Left to right: Chancellor Larive, Fahim Hasan Khan (holding award), Donnie Stewart (me), Vice Provost Biehl
Secure Message System
2021
Developed a Secure Messaging System that implements and combines many important cryptographic techniques including Hashing, Signing & Authentication, Key Ratcheting, Encryption & Decryption, and the Block Cipher Mode of Operation.
​
The code, how to run, and technical details all can be found on my GitHub. This diagram shows the full pipeline required to send messages back and forth.
Sign Language Detector
2021
American Sign Language (ASL) is a complete, complex language that employs signs made by moving the hands combined with facial expressions and postures of the body.
​
For this project, I focus on classifying these hand gesture images using Convolutional Neural Networks. Also, I compare transfer learning models and models trained from scratch. Technical details and Colab Notebook are accessible on my GitHub.
The chart shows my Bitcoin Price Prediction (y_pred_avg) plotted next to actual Bitcoin Prices (y_test) taken from 529 random dates from the last 7 years.
Bitcoin Price Predictor
2021
Bitcoin, amongst other assets like AMC and GME, has been all the rage this past year and reached an all-time high of $66,974 per bitcoin.
​
The purpose of the project is to explore using various Recurrent Neural Networks (RNN) supported by TensorFlow, on tabular time series data. Specifically, the task at hand is to predict the next day's high price of Bitcoin based on prior closing date prices. Technical details and Colab Notebook are accessible on my GitHub.
Analysis of Twitter's Trending Topics
2020
Analysis of 2011 Twitter's Trending Topics in Comparison to Top News Headlines. The project contains code that processes and visualizes trending Twitter data based on a labeled dataset.
​
The table shows a screenshot from the interactive Google Facets visualization library. This interactive table of the Twitter data can be found on my Github.
