Joseph Olufemi Davis

I'm a rising sophomore at Howard University in Washington, DC, interested in computer vision and perception. I am dedicated to pursuing a PhD, doing research on machine learning and CV. At Howard, I conduct research with Dr. Danda B. Rawat in the CoE-AIML.

I am currently looking for research opportunities for next Summer (2027).

Email  /  CV  /  LinkedIn  /  Github

profile photo

Research


I'm very interested in object detection and scene reconstruction. I am excited for future research work, and hope to continue to produce publications throughout my undergraduate career and beyond.
Comparative Study of Point Cloud, Voxel, and Mesh Representations for 3D Semantic Object Detection
Joseph O. Davis, Danda B. Rawat
SPIE Defense + Security, 2026

The mesh data representation is more accurate, but considerably slower than the voxel and point cloud representations. Thus, it is more appropriate for less urgent tasks. The voxel representation is the fastest, but the least accurate. Thus, it is better suited for tasks where speed is more critical.

Projects

Diabetic Retinopathy Predictor
Python, PyTorch, Scikit-learn
Repo

Developed a machine learning model that predicts diabetic retinopathy severity (1-5) from retinal images using PyTorch. Built a custom, robust CNN architecture to classify retinal images with 0.80 Precision, 0.82 Recall, and 0.80 F1 Score. Aggregated and preprocessed a dataset of 30,000+ retinal images from Kaggle to train the model.

Sign Language Letter Classifier
Python, Tensorflow, scikit-learn, Keras, OpenCV, MediaPipe
Repo

Created a dataset with 12,000+ hand gestures of ASL letters using Google's MediaPipe framework. Trained a quick MLP on the data, allowing for real-time inference through video footage.

Jam Plan Scheduler
Java, CSV
Repo

Designed and implemented a scheduling algorithm that optimizes class schedules for 700+ students based on their course preferences and availability. Deployed the algorithm to assist academic advisors with course planning, resulting in improved student satisfaction and reduced scheduling conflicts


Design + Source Code from John Barron's website.