Skip to content

Generative 3D model for simulation-based inference of a warehouse recognition system.

Notifications You must be signed in to change notification settings

srad/Warehouse.Unity

Repository files navigation

Warehouse.Unity

This repo was part of a System Engineering course at the Goethe University Frankfurt/Main in which the goal was to implement a AI palett detection system for a moving forklift.

For this purpose, this repo is the implementation of a probalistic world model which provides a simulation and training data for the AI model. It is based on Unity as a generative farmework and implemented in C#.

The entire system consists of two part implementation which generates training data (this repo) and further processes the training data and finally generate a hypothesis.

training

Overview of the world model

This is a high level overview of the warehouse world model in which the forklift operates.

world_model

Material model for the paletts in the warehouse

material_model

Unity editor view

This view allows to setup simulation parameters for the generative model. The entire warehouse can be dynamically regenerated

unity

Top view of a scene

scene_top

Annotation hints

The colored materials are for illustration how every part can be tracked for training data generation with annotation meta data.

pallet_colors

About

Generative 3D model for simulation-based inference of a warehouse recognition system.

Resources

Stars

Watchers

Forks