Control of a video-game race car using a convolutional neural network
In this project we tackle CarRacing-v0, OpenAI Gym’s 2-D racetrack, treating each rendered frame as an image-classification task. By converting raw pixels into steering and throttle commands, we train an agent that masters the course—demonstrating how modern computer-vision pipelines and reinforcement learning can turn a blank-slate neural network into a self-driving racer.
https://messi-orion10.github.io/s_orion10.github.io/files/HW2_ORELLI.pdf
