Crowd Sourced Satellite Image Tagging for Machine Learning (Ref 17/20)

Company: Satellite Applications Catapult, Applied Digital Intelligence Team

Location: Harwell, Oxford

Background: Satellite image analysis can be a complex and time-consuming process. In the last half decade, automatic image identification algorithms have drastically improved, although they do not yet achieve the same quality of results generated by human operators. To increase the reliability of the automatic image recognition output, the algorithms need to be able to access thousands of annotated images to use as training data. Whilst there are large repositories of annotated images (e.g. cats and human faces) taken from conventional terrestrial photography, there are only limited annotated catalogues consisting of satellite imagery.

Recently there have been great success stories using the power of the public to annotate large amounts of data, which would otherwise be difficult to achieve through a single person or organisation. Examples include Zooniverse, a crowd sourcing company that made its fame with its Galaxy Zoo application, allowing the public to classify thousands of astronomy images. By using similar crowd sourcing techniques, it is possible to create a diverse training dataset for automatic image recognition.

The Satellite Application Catapult is in a unique position where it has access to the Sentinel 1 and 2 data via its cloud computing facility CEMS. This represents an opportunity to create a Sentinel-dedicated crowd sourcing platform to train and validate image recognition algorithms.

This project would be a follow-on from last year’s SPIN Internship, taking the web application and embedding it onto an iOS or Android platform.

Project Description: The Catapult is looking for enthusiastic and skilled students to begin to develop a mobile-based crowd sourcing application utilising the Catapult’s Sentinel data access.

Student Specification:

Applicants will need to be numerate, have computing skill, be able to critically evaluate problems, suggest solutions and show initiative in a supervised R&D project.

  • Experience in one or more of the following areas: application development, computer science, geography, remote sensing and maths.

Target courses: IT, Web development, Computer Science, Physics, Remote Sensing, Engineering, Physical Geography, Mathematics and other similar courses.

The Nitty Gritty: 8 weeks fixed-term contract to be agreed with successful candidate but nominally with a start date around 26 June which is also the SPIN Induction day at Harwell. Salary is £1,500 per calendar month. (guide only)

Closing Date for Applications: 24 April 2017. Applications will be through the online form attaching a CV, before the closing date. They will be checked for eligibility and forwarded to the employer. Please include the reference number in the email subject line.

Fill out my online form.

Cookies on Catapult explained

To comply with EU directives we now provide detailed information about the cookies we use. To find out more about cookies on this site, what they do and how to remove them, see our information about cookies. Click OK to continue using this site.