Satellite image annotations for key items recognition

Airbus in a few words 

• Airbus SE is Europe number one and the World number two in the aerospace and defence industry.
• Founded in 1970, Airbus currently employs more than 130,000 people.
• Airbus Defence and Space (ADS) is one of the three divisions of Airbus Group, specializing in military aircraft, drones, missiles, space launchers and artificial satellites. 
• The OneAtlas platform, which provides quick and easy access to quality satellite imagery, layers and analysis for a wide range of use cases.
  

Isahit in a nutshell:

• First ethical data labeling platform for AI and data processing in Europe
• 2000 projects with over 1700 HITers
• HITers established in over 39 countries
• 300+ satisfied customers
  

Airbus defense and space mission: 

Democratize access to satellite imagery to help users solve the challenges of our world, such as:

• Economic
• Environmental
• Defense and security
• Food supply
  

Today, artificial intelligence can automatically detect objects and changes between satellite images.

OneAtlas Data, the platform used by Airbus Defense and Space

One Atlas Data is a platform for quick and easy access to high-quality satellite imagery, layers and analysis for a wide range of use cases.

OneAtlas Data provides access to a combination of high and very high optical resolution with SPOT and Pleiades. Both offer daily revisit capability, ensuring efficient monitoring.

Use case example: Satellite imagery dataset for aircraft detection

Image: 

This worldwide training dataset consists of more than 25.000 annotated aircrafts on 0.5m.

It includes:

• Small aircrafts
• leisure 
• private jets 
• regional commercial
• Large commercial
• Helicopters
• Fighter jets
• Bombers
• Other military

Artificial Intelligence model trained on this will detect aircrafts on 0.5m resolution images with a performance of:

> 85% in detection rate 

> 90% in precision rate

Challenges encountered

- Experts are not available and data scientists should not annotate all data 

- Need for massive and consistent annotations in a short time

- Quality must be assured

To obtain massive and consistent annotations in a short time, Airbus chose crowdsourcing, which allows to outsource the workload to hundreds of remote contributors. 

The solution: outsourcing with impact thanks to isahit

In order to meet these challenges, isahit appeared to be the most appropriate solution as it offers : 

• More than 1000 active contributors in 39 different countries, able to respond to annotation needs in a short time
•  A very demanding quality review:

Isahit implements a very strict project monitoring, throughout its realization, trains these contributors and implements tests upstream to ensure that a constant quality is maintained.

• Great agility and ability to adapt resources to changing needs
  

In addition, Airbus shares the same values as isahit and supports its mission: to make tech and AI a work and training opportunity for women around the world.

A DOUBLE VICTORY: TECH & SOCIAL

Two years into our collaboration, we are mutually excited to report the following: 

• 500,000 annotations made by our contributors: Bounding boxe oriented / Bbox classic / Polygon
• On the flow project
• 55 HITers worked on the project (self confidence, money to finance their project, digital skills) 
• 53h of training provided (specific labeling tools but also nlp projects, google sheet, google doc, linkedin)