Directors: Sisi Aarukapalli & Pranav Nair
Led by Aarian Ahsan
You will learn the basics of image classification to then be immediately thrown into the ringer to train a model that can detect certain types of objects under various measures of turbid water by utilizing CNNs (Convolutional neural networks) such as YOLO or MobileNet. The main goal is to be able to demonstrate underwater image classification at a high accuracy rate. (If all goes well we’ll be able to see it in action with a live ROV from Robosub).
Led by Chris Back
You will attempt to fool machine learning models trained to detect political disinformation on Twitter using adversarial learning. An adversarial example feeds a model seemingly healthy input which is actually designed to deceive the model into outputting an intended mistake. This project aims to identify if political actors (ex. Russia-Ukraine Conflict) can bypass detection systems using adversarial learning to spread misinformation to fool public understanding of social media.
Led by Ben Bowers
Images that are used as training data for machine learning are often more effective if they use a specific color scheme, such as thermal coloring. Using CycleGAN, we will build a model that can convert a given photo from normal coloring to specific colorings that are more effective as training data.
Led by Rohan Dave
Blockchain technology is as popular as ever due to the usefulness of the infrastructure it offers. However, the backbone of this versatility, smart contracts, can be dangerous if incorrectly created. In this project, you will learn about Ethereum blockchain and smart contracts, and jointly analyze commonly used smart contracts on the Ethereum network in an attempt to find vulnerabilities or logical flaws that we can manipulate to our advantage.
Led by Aditya Desai
You will learn the basics of deep learning and data augmentation to develop a model that can accurately detect a wide range of diseases from a Chest X-Ray. Some diseases that will be focused on include: Pneumonia, COVID-19, Cardiomegaly, and a healthy thoracic cavity.
Led by Lawson Lay
Through the use of consecutive processing of Neural Radiance Fields (NeRFs) using bleeding-edge NVIDIA technology, you will learn, develop, and optimize a process to create a video which captures a 3D Virtual Reality of that moment in time. We'll go over the basics of NeRFs, different video creation methods, and the limitations of commercial CUDA cores. Our end goal is to develop a relatively efficient process of recording VR videos using NeRFs and to present a room-scale dynamic VR NeRF video.
Led by Naveen Mukkatt
Reinforcement learning algorithms are responsible for some of the most spectacular breakthroughs in many games, such as Chess, Go, Starcraft II, and Stratego. In this research project, you will learn about reinforcement learning algorithms, and then implement your own reinforcement learning algorithm to teach two agents to play a game similar to Tag or Capture the Flag in the Unity game engine using its ML Agents extension.
Led by Chris Sheppard
Diagnosing wrist fractures requires time-consuming specialized training. We will use leading computer vision algorithms for automated fracture detection.
Led by Adith Talupuru
Computing the transformations of rigid bodies over 3 dimensions is a very intensive process. A lot of computers have quite a bit of difficulty doing it efficiently. It's the main reason graphics cards exist in the first place. Let's try making it work on an embedded device!
Led by Rishit Viral
The Internet of Things (IoT) has brought about a revolution in connected devices, making our lives more convenient and efficient. However, as more and more devices are connected to the internet, the risk of security vulnerabilities increases. In this research project, we will be focusing on identifying security vulnerabilities in IoT devices and exploring potential solutions to address these vulnerabilities. By finding and addressing these vulnerabilities, we can improve the overall security of the IoT ecosystem and protect against malicious attacks.