Giter Club home page Giter Club logo

samm's Introduction

Semantic Aware Map Metrics

This is the repository of the paper Reconstruction of 3D Semantic Map and Its Quality Estimation.

Here are several tools for calculating quality metrics to reconstruct 3D surface maps with/without semantics described in the paper. There are also several scripts for visualizing map quality.

Here is an example comparison of the quality of two different maps in the same places:

Comparison of the quality of two different maps in the same places

Metrics

For map evaluation, the following metrics are used:

  • Precision

  • Recall

  • F-Score

  • Chamfer Distance

  • Hausdorf Distances

The mathematical calculations are as follows

Basics

The quality evaluation of a 3D map comes down to evaluating the proximity of two point clouds. LiDAR data (randomly selected from) is used as gt, points randomly selected on the map are used as "Map points".

The shortest distances between the point clouds are used as the base values for calculating the metrics. Namely, the following values are calculated:

  • Array of shortest distances from map points to the gt points $$D_{forward} = d(m,G_l) d(m,G_l) \gets \min_{g\in G}|m-g|_2$$

  • Array of shortest distances from gt points to the map points $$D_{backward} = d(g,M_l) = d(g,M_l)\gets \min_{m\in M}|m-g|_2$$

Metrics itself

  • Precision $$P = \frac{100}{|M_l|}\cdot \sum_{m\in M_l}\left[ d\left(m,G_l\right) < d \right]$$

  • Recall $$R = \frac{100}{|G|}\cdot \sum_{g\in G}\left[d\left(g,M\right) < d\right]$$

  • F-Score $$F = \frac{2\cdot P(d)\cdot R(d)}{P(d)+R(d)}$$

  • Chamfer Distance $$CD = \frac{1}{2|G|}\cdot \sum_{g\in G}\left(g,M\right)^2 + \frac{1}{2|M|}\cdot \sum_{m\in M}\left(m,G\right)^2$$

  • Forward Hausdorff Distance $$HD^{forward} = \max_{m\in M}d\left(m,G\right)$$

  • Backward Hausdorff Distance $$HD^{backward} = \max_{g\in G}d\left(g,M\right)$$

Each metric is calculated separately for each semantic class.

More details, requirements and HOW-TO will be available soon

samm's People

Contributors

vitalyybezuglyj avatar

Stargazers

 avatar

Watchers

 avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.