Projects

For now, the information here will largely be taken from the Github repositories for these projects, which are publicly available. I will also be adding more information on the research I’ve done outside of projects soon!

Eco-Civilization MDP: A Multi-Agent Environment for Sustainable Growth

Eco-Civilization MDP is a reinforcement learning project that adapts the Civilization game to explore sustainable development strategies. Utilizing Multi-Agent Proximal Policy Optimization (MAPPO), the project trains AI agents to balance economic growth with environmental responsibility. The modified game introduces environmental penalties, challenging agents to optimize strategies that mitigate climate impact while expanding their civilizations. With the work, our aim is to use reinforcement learning for finding strategies for enabling countries to stay competitive economically while taking environmentally conscious actions.

FocusCaption: Image Captioning Utilizing Saliency Models

This project aims to implement an image captioning method that utilizes 2 CNNs in parallel with a saliency prediction model to extract extra information from where “should be focused” in an image. For this implementation, the saliency model being utilized is TranSalNet. After figuring out the location that should be focused on, the portion is cut out and fed into an identical CNN, and at the end, the information from the full image and the focused portion is combined and fed into a decoder.