This project is split to multiple environments. In this fork of the project we initialized the Tiago environment (see folder Tiago). This environment has multiple scenarios
- the empty tabletop scenario called BuildWorldScenario in the module build_world_learn_grasp.py
- the tabletop scenario with the big static object in the middle called BuildWorldScenarioStatic in the module build_world_static.py
- the tabletop scenario with the small table on top called BuildWorldScenarioTable in the module build_world_top_table.py
On the other hand there are the deep learning models that were trained differently and compared to get the best results in folder deep_heuristic/models. The folder deep_heuristic also has the data generation tools, some utils, debug and visualisation tools.
To experiment with the models please navigate to deep_heuristic/models and run the module exp after loading some data drom the nn_utils module or a depth image of your choice.
- Draw a minimum sphere around the target object.
- Sample points on the sphere by single-sphere icosahedral discretization.
- Graph convolutions from a higher resolution grid to a lower.
