Custom ML Algorithm Development for Object and Feature Change Detection
Earth-i works with companies and government agencies around the word to create custom multi-resolution and multi-sensor data-agnostic machine learning algorithms to answer their specific questions at scale. We are engaged with numerous customers including space agencies and commercial businesses to understand their business objectives and match these to data sets and machine learning methods.
Approach
We pre-agree an achievable target accuracy and define the inputs and outputs rigorously to ensure that these are met.
Once completed we then transfer the Machine Learning algorithm to the customer’s infrastructure which enables them to create an automated processing chain for operational use, and provide ongoing support, development, and training.
Support offered
We assist customers with varying levels of remote sensing, infrastructure, and Machine Learning knowledge to achieve their aims through:
• Programme development and management
• Project definition and Algorithm and System specification
• Data procurement, data labelling and ML model training
• Machine Learning model testing
• Machine Learning model transfer and operationalisation
• Ongoing maintenance and support
• Customer training
Usage
Our customers have used our technologies for object detection and building footprint detection to inform national planning regulations. We have extended these technologies for other applications such as crop detection and agricultural monitoring. We have also used our expertise to help with defence and intelligence work, for example to help identify military airframe assets, at scale, around the globe.
Data sources used
Data sources include video, optical and radar data allowing these capabilities to span multiple domains and atmospheric conditions.
