Estimating Bicycle Route Attractivity from Image Data

TL;DR

This thesis develops a CNN-based method to estimate bicycle route attractivity using street view images and geographic data, aiming to improve road scoring accuracy for cycling infrastructure planning.

Contribution

It introduces a novel approach combining street view images, map data, and transfer learning to enhance bicycle route attractivity estimation.

Findings

Enhanced dataset improves model accuracy.

Transfer learning reduces overfitting.

Various architectures tested for optimal performance.

Abstract

This master thesis focuses on practical application of Convolutional Neural Network models on the task of road labeling with bike attractivity score. We start with an abstraction of real world locations into nodes and scored edges in partially annotated dataset. We enhance information available about each edge with photographic data from Google Street View service and with additional neighborhood information from Open Street Map database. We teach a model on this enhanced dataset and experiment with ImageNet Large Scale Visual Recognition Competition. We try different dataset enhancing techniques as well as various model architectures to improve road scoring. We also make use of transfer learning to use features from a task with rich dataset of ImageNet into our task with smaller number of images, to prevent model overfitting.