ROS Resources: Documentation | Support | Discussion Forum | Service Status | Q&A answers.ros.org

November ROS-A Europe Community Meeting

ROS-Agriculture
#1

The next ROS-A Europe meeting will be 2020-11-26T13:00:00Z

Tentative agenda:

  1. A presentation given by @Nikos_Mylonas- Eden Library: an initiative to address the challenges of agrifood data scarcity and boost AI vision tasks
  2. Extending the simulation of lawn tractor demo for a hackathon
  3. Open discussion

Join Zoom Meeting

Meeting ID: 509 011 0612

Best regards,
Karen

3 Likes
#2

ROS-Agriculture Europe Meeting Minutes 2020-11-26

Attendees

Name Organization
Ting-Chia Chiang TUDelft
Ilias Amri Sorbonne Université
El Houssein Chouaib HARIK NIBIO
Nikos Mylonas AUA
Ioannis Malounas -
Ragesh Ramachandran Fraunhofer IPA
Avgoustakis Giannis -
Tim Clephas -

Topics

  1. A presentation given by @Nikos_Mylonas - “Eden Library: an initiative to address the challenges of agrifood data scarcity and boost AI vision tasks” (slides (934.3 KB))
  2. Extending the simulation of lawn tractor demo for a hackathon
    • The goal is to set up a challenge like DARPA subterranean challenge
      • Have a virtual environment for navigating in row-crop fields - possibly base our work on a project from University of Bonn
    • Topics can be covered:

      • Data collection and training pipeline:

        • Train data for plant recognition from synthetic data (taken from photogammetry). Details can be found in this post

      • Localization:

        • There is still a big challenge in outdoor localization, especially in GPS-denied environments
        • Ioannis is working on a visual slam project using RTAB-Map
        • Configure a receiver to connect to a network of RTK bases to increase localization precision- Centipede project

Recording

The video of this meeting may be found here

1 Like
#3

Regarding setting up a challenge, there’s the Field Robot Event that’s already planning an online competition this year. They’re a yearly contest for student teams that’s usually held somewhere in Europe. Tasks involve autonomous (non-GPS) row navigation, and usually something that has to do with weed detection/removal. If you want to contact them, Prof. Griepentrog probably is the right person.

1 Like
#4

Hi everone,
I just found out this thread and the presentation which was quite interesting based on work that I have done in the past as well. I have a few comments and want to get upto speed:

  1. Regarding the farming simulator, being a source of synthetic training data, I did something similar using GTA 5 video game for LiDAR data using ray cast algorithm in my master thesis. The code repo is here, (github repo)[GitHub - niqbal996/Lidar_gta_5: A LiDAR plugin for GTA 5 that generates data similar to KITTI data set.]. The application was intended for autonomous car driving and not the agriculture plant detection. But the work is similar. Is the farming simulator with the ray cast based LiDAR sensor public? Can someone point me to the link? Thank you.

  2. I have already talked to @Ilias Amri before and working on the synthetic plant data generation and there I can also contribute to the baseline with a real-world dataset (Sugarbeet 2016 dataset) hopefully soon with a basic machine-learning pipeline.

2 Likes