CropCraft: A Procedural World Generator for Robotic Simulation of Agricultural Tasks
Researchers have developed CropCraft, an open-source tool built with Blender and Python to generate realistic 3D agricultural environments for robotic simulations. This tool can create diverse scenarios like intercropping and vineyards, complete with detailed plant models and realistic spatial variability. CropCraft also provides ground-truth annotations for elements within the generated worlds, facilitating the development of perception and navigation algorithms. Initial applications show that models trained solely on CropCraft-generated data achieve a competitive sim-to-real performance for crop-weed semantic segmentation. AI
IMPACT Enables more robust training of agricultural robots by providing diverse and realistic synthetic environments.