Probabilistic Brush Compression
A new image compression algorithm I designed, currently deployed on Hugging Face Spaces with an interactive Demo.
Active Perception with VLM
Project applying multimodal language models for active perception robotics tasks. Done at the Institute of Intelligent Systems and Robotics (ISIR)
Ege's Archive Indexation App
A media archive indexation app for photographers, with many features for convenient archive browsing, selection and file system management.
PersonaSFT
A pipeline that takes in various social media data exports from you (discord, whatsapp, etc), and trains an AI model to talk like you.
TaMtD: Compressed-Domain Visual Understanding Across Codecs
A research project using multi-stage decoding to isolate and measure the accuracy-preprocessing trade-off between intermediate and terminal codec representations.
SifuTAS
A mod with some TAS tools for Sifu
Image Generator
Locally trained GAN model that generates photographs of birds from sketches. Used a full MLops stack: kaggle, pytorch, docker, mlflow, gradio and hf spaces.
AI Chatbot
An AI Chatbot LLM model with architecture based on the original GPT2 (124M), built and trained entirely locally from scratch.
Lunar Lander AI
A reinforcement learning project for the Lunar Lander environment, simulated on the OpenAI Gymnasium library, trained using EA, DDPG and TD3 algorithms
Multi Agent Coordination
Optimizing the communication and coordination protocols between multiple agents in a shared environment. Built using the JADE/DEDALE framework.
AI Image Upscaler
A web interface built with Flask, HTML, JavaScript and CSS that uses Hugging Face transformer models to upscale the given images.
Multimodal Chatbot Voice Interface
A Gradio interface for multimodal voice interaction with local vision language models.
Detect Object
Detects the object in an image and crops out the rest. Purely algorithmic approach of my design, not using AI or external libraries.
Simplify Color
I developed an image quantization algorithm, then did a survey comparing results from my algorithm with the industry standard, results are in the repo.
Movement Tracker
Tracks the movement in a given video, placing a red dot on the most active object. Algorithmic approach, no AI used.
Submerge
Simulates and visualizes an area of land being submerged based on the given topographical data and satellite image.
AI Number Reader
An AI program that can be trained to read numbers on a 16x16 black and white image.
AI Robot Simulation
An AI based solution for a simulated robotics project in my "AI and Games" class
Smaller Fish
My submission to the Pygame Community Summer 2023 Game Jam.
Simplify
Creates a simplified version of a given image by finding the high contrast points.
Simplify Video
Creates a simplified version of a given video by finding the high contrast points.
Flies and Flowers
Simulates a few flies and different types of flowers on a flat terrain. Project for a java class