Hi there!
I started to work on randomly generated Ignition Gazebo test environments for agriculture robots, and Iām so happy with the first results that I wanted to share it with you 
Briefly, in the current state:
- a Blender script generates the crop meshes (tomato only), and a Jupyter notebook puts together the SDF models and world (based on the given parameters).
- It generates ground truth data (positions of plants and fruits) for each world and model, which can be used to evaluate the robot performance.
See the readme for more detail, instructions to run it in Docker or in a Catkin workspace, generated example worlds, etc.
PS Iām looking for a job in robotic agriculture. If you think you might want to work with me, please send me a line (linkedin)