farming revolution tensorflow-and-ros-mini-project

Mini-project to evaluate capabilities in tensorflow and ROS for applicants at farming revolution GmbH

Objective

The objective of this mini-project is to evaluate an applicant's

• capacity to use linux
• basic understanding of ROS and tensorflow
• basic programming skills in python
• capability to use git
• ability to write proper documentation

Preliminary tasks

• Install ROS and tensorflow.
• Recommended: Beginner Level tutorials 1-7 and 12 of ROS: link to ROS tutorials.
• Recommended: Customization basics: tensors and operations tutorial for tensorflow: link to tutorial. Depending on your background, starting with an additional beginners tutorial might help as well.

Play the bag file

The provided bag file contains images of plants published on the /img_bgri topic (4-channel image, blue, green, red, near-infrared).

rosbag play mini-project.bag -l
rosrun image_view image_view image:=/img_bgri  # Visualize the image

Task

Write a python ROS node, that uses tensorflow to extract a basic plant mask.

Your node has to

• subscribe to /img_bgri
• publish to /plant_mask a mask (single-channel) showing the plant pixels

Do not write something too complex: the goal is prove that you can write a simple ROS node with minimal functionality. An easy way to detect the plant pixels is to subtract the red from the near-infrared value and apply a threshold.

You have to use tensorflow to do that (even if it would be easier with an OpenCV or numpy function).

Morphological operations can be used if you want but it is not required.

Upload on github.com

Use a repository management service supporting git (preferrably github.com) to:

• Create a private repository
• Push your code
• Add a README.md file describing how to install the required dependencies, what your node does an how to use it
• Send us a link to the repository and give us permissions to read your code.