Vehicle Detection In The Wild
A freelancing project to detect vehicles in the wild by processing aerial imagery taken from drones.
Different computer vision models where used. The final model was based on the YOLO model.
Data annotation was done using OpenCV CVAT Pytorch was used for modeling.
Sickness Xray Based Detection
Used several convolutional neural networks CNNS model to predict the sickness of patients.
Tried Alexnet, Resnet, Densnet, and several custome models. The work opened a research oppurtinuty to explore
how such models not only predict sickness but also detects the patients protected charecteristecs like sex, age,
and race. We found that although networks can predict such charactersitics but they doesn't baise on them while
predicting sickness.
Github
Text Segmentation Baseline
N-gram based machine learning text segmentation applied over islamic hadith to predict the split
between sanad and maten parts. Islamic ahadith are historical narrations that describe the acts
and statements of Prophet Mohammad (PBUH), his household, and his Companions. Formulated a new
entropy based segmentation technique. Used Python, Requests,
BeautifulSoup.
Github
Dubuque City Traffic Monitoring
Developed a full data application that routinely featchs data from the clients API and dump them processed in database.
Different machine learning algorithems (LSTMS, SVMs, Knns, Kmeans,
...) are then executed and the resulting models are finally served using Flask. A unifiying dashboard was developed
to monitor the data, visualize statistics anad ML summaries.
Unsupervised Deep Learning
Developed a research framework to understand the representation quality of several Unsupervised Deep Learning Models. We are talking
about models that learns without lables using clustering, entropy, rotation, masking to generate from-data groundtruth. The research also
tackled how to optimize them from efficancy perpective without trading off performance. The research was acumulated to serve as my master dissertion.
Github
Point In Polyhedron
A CUDA implementation of the Point In Polyhedron (PIP) problem. PIP is an important problem that have a
number of applications in different domains such as computer graphics. In this project. I combine and
re-implement two research papers trying to parallelize the procedure on GPU using
CUDA C.
Github
Kure
Lead and contributed to the development of kure.
Kure is a fully containerized application that allows users to record their voices which is then transcribed and analyzed using AI.
Used Flask, Docker, Django, Postgres, Prometheus,
and Grafana. Used three different AI models to transcribe and analyze voice notes.
The app is developed with CI/CD in mind.
Github
>
