Template rendering for ZS6D: Zero-shot 6D Object Pose Estimation using Vision Transformers

1. Set up conda environment:

conda env create -f env.yml

conda activate render_temp

Rendering with bop datasets

2. set up the config file

bop_templates_cfg.yaml

3. run script to render templates

blenderproc run render_bop_templates.py bop_templates_cfg.yaml

Rendering with custom mesh files (ply/obj)

2. set up the config file (mesh files are expected to be in mm)

mesh_templates_cfg.yaml

3. run script to render templates

blenderproc run render_mesh_templates.py mesh_templates_cfg.yaml

Pose generation

If more poses than 2562 are wanted, further poses can be calculated with the script below In line 133 for the script increase the levels, higher means more poses

blenderproc run generate_poses