IGVC Net

Introduction

IGVC Net is the convolutional neural network (CNN) used for the IGVC competition to detect white lines on the course. This page documents how to train and evaluate IGVC Net using pytorch, and how to load a model for running the vision ROS node.

Requirements

• python 2.7 (should already be default w/ ROS installation)
• CUDA 9.0 (for GPU acceleration)
• pytorch 0.4.0 (https://download.pytorch.org/whl/cu91/torch-0.4.0-cp27-cp27mu-linux_x86_64.whl)

Training/Evaluation

Dataset setup

1. Download the example data

TODO