Source: Deep Learning on Medium
The Project — AutoBound
I will be working on building a plugin for JOSM that can automatically identify the rooftops of buildings in a given map area. This will make it easier for contributors of OSM to easily mark buildings and add data.
The tool will have two parts:
- The Front End — This will be written in Java and will act as an interface between the user and JOSM. The plugin will allow users to select an area on the map. The satellite image of this area will be Base64 encoded and sent to the Back End.
- The Back End — This part will be written in Python. The Back End will have endpoints listening on a port (default: 5000). When it receives data from the Front End, it is decoded and processed. Image Segmentation is done by a Deep Learning model that has been trained to segment rooftops in an image. Then the position of the nodes that can create a boundary for that rooftop is calculated and is returned to the Front End, where it is displayed on the Map View to the user.
The Back End will also be made available as a Docker image to make it easier to use.
GSoC gives students three months to complete their project.
For the first month, I will be working on the Front End of the Plugin. I will be pushing all my code to the GitHub repo every day.
For the following two months, I will be collecting and preparing data to train the Deep Learning model, training the deep learning model, integrating it with the endpoints, testing the working of the entire stack and adding documentation. The code for the server will be pushed to this GitHub repo.
I will also be posting updates on my work every week on Medium.
Till my next post, Good Bye 😃