Data Collection Overview
Proper data collection goes a long way towards making calibration easier and more robust. The right methods ensure better results while minimizing the total amount of data needed; in other words, good data gives us more for less.
The guidelines specified here are not hard requirements to use the Tangram Vision Platform; they're just guidelines. Different projects will have different constraints, and it's not always possible to do everything here.
However, these guidelines have largely been designed to produce consistent and reliable calibration processes when calibrating with TVCal.
Advanced Topic: Projective Compensation
Most data collection strategies are specifically aimed at reducing projective compensation wherever it may occur. Learn more about this term in the docs.
Component-Specific Calibration Strategy
What makes a data collection process "good" depends on the component configuration in a system. This makes it difficult to make general guidelines. However, there are a few key tips to follow:
- Model all systematic effects
- Reduce projective compensation (through modeling or by collecting more data)
- Increase observation precision
Specific guidelines for different components can be found below.
Currently, the Tangram Vision Platform only supports camera components. We appreciate your patience as we add more components to our list of supported modalities.
|Camera||Camera Data Capture|