Project repository for COE 379L: Software Design For Responsible Intelligent Systems
Spring 2024, UT Austin
The project focuses on applying exploratory data analysis (EDA) and machine learning techniques to predict car fuel efficiency. It uses various Python libraries such as pandas for data manipulation, seaborn and matplotlib for data visualization, and scikit-learn for model implementation.
Project 2: Cancer Predictions (Classification)
This project aims to develop machine learning models to predict the recurrence of breast cancer in patients. The dataset used contains variables such as age, degree of malignancy, tumor size, and others. Various supervised learning techniques are used to create ML models to predict whether a patient will have recurrent breast cancer. The models focus on minimizing false negatives to ensure high recall, recognizing the importance of accurately identifying potential recurrences.
Project 3: Damage Prediction (Image Classification)
This project focuses on using deep learning techniques to classify satellite images into "damaged" or "non-damaged" buildings after Hurricane Harvey. The dataset used is a collection of satellite images from Texas. This project included data preprocessing for image analysis, exploration of various neural network architectures such as ANN, LeNet-5, an alternative LeNet-5 architecture, VGG-16, ResNet, and Xception and also included evaluation of model performance. The models are deployed with Flask , providing an HTTP interface for image classification.
Project 4: Dog Breed Prediction (Image Classification)
This project focuses on predicting the breed of a dog (or the potential breed of a human) from pictures of dogs (or humans) using TensorFlow neural networks. We explored various neural network architectures, deploying the best-performing model with a Flask-based inference server. This README covers the deployment and usage of the model inference server. This model can predict 120 different dog breeds, as feature in the Stanford dogs dataset.