While we have repositories and rudimentary procedures for testing the following:

1. semantic segmentation modeling
2. live mapping
   ... we do not have a comprehensive, reproducible framework for testing all aspects of the system.

We will need a framework that can provide benchmarks and comprehensively run tests (with the correct metrics) for the following:

1. Semantic segmentation modeling
2. Object detection algorithm
3. Accessibility feature Localization and Attribute Estimation

For point 3, we will need to have dedicated benchmarking datasets of outdoor mapping, that can compare different implementations. Currently, we do not have that.