High Precision Leaf Instance Segmentation for Phenotyping in Point Clouds Obtained Under Real Field Conditions

This repo contains the code for our publication "High Precision Leaf Instance Segmentation for Phenotyping in Point Clouds Obtained Under Real Field Conditions". The paper can be downloaded here.

Setup

• Install docker, docker-compose and nvidia-docker2 on your machine
• Clone this repo to a folder on your pc
• Edit docker-compose.yml to set /data and /logs paths. /data contains the dataset and /logs is the folder where the tensorboard logs will be saved. Edit the part of the paths before the ':' to make it point to the directories on your machine.
• Execute make build
• Get a coffee and wait

Train

• To train the leaf segmentation model with the provided config file run make train_instances CONFIG=config/leaf_segmentation.yaml

Test

• Execute make test_instances to evaluate the segmentation network

Data

This approach has been tested on sugar beet and tree point clouds and has shown good performance. The data used in our paper is not publicly available at the moment but will be in the future.

Citation

If you use this repo, please cite us :

@article{marks2023ral,
    author={Marks, Elias and Sodano, Matteo and Magistri, Federico and Wiesmann, Louis and Desai, Dhagash and Marcuzzi, Rodrigo and Behley, Jens and Stachniss, Cyrill},
    journal={IEEE Robotics and Automation Letters},
    title={High Precision Leaf Instance Segmentation for Phenotyping in Point Clouds Obtained Under Real Field Conditions},
    year={2023},
    volume={8},
    number={8},
    pages={4791-4798},
    doi={10.1109/LRA.2023.3288383}}