TensorFlow, Object Detection, Lancaster City PA Row Houses, Social Ecology

This project is a test to try out machine learning using TensorFlow Object Detection Models. If a TensorFlow model can be trained to detect row houses, can it further be trained to detect row houses of varying styles, and if so can the detection model eventually tell us about the visual makeup of different areas within a city? There have been a few conceptualizations of neighborhood definitions in Lancaster City PA over the past 40 years or so (link). Is it possible to couple object detection methods from machine learning with individuals' own ideas of their neighborhood boundaries (link) to create a visual representation of neighborhoods as the spaces that the individuals within them define them.

TensorFlow Object Detection Setup

The following tutorial was used to implement the object detection model in TensorFlow: Tutorial Link

Getting Images to Train the Model

The following section describes the process of using the Google Street View API to collect images. 500 images were gathered from Google Street View API with the following method