CPS Early-Stage Detection and Control of Leaf Diseases in Tomato Transplant Production: Mini T-REX tomato image dataset

The goal of this project is to develop an integrated Cyber Physical System (CPS) for early plant disease detection and control based on precise pathogen identification in transplant greenhouses. The system consists of three components: 1. a robotic platform to collect images and leaf DNA samples with possible pathogen infection. 2. a microfluidic device for on-board DNA sample extraction of all microbes, including the putative pathogens, and identification of pathogen by nanopore sequencing in near real-time. 3. a method of detection of disease spread using imaging and sequencing data, and to determine optimal control strategies. This dataset include tomato plants RGB images and depth data for 3D reconstruction using a robotic platform called Mini T-REX developed for this project.

Who: Data were generated by a GRA from Virginia Tech

What: Multi-view RGBD images and longitudinal RGB images of plants (tomatoes, lettuce, and pennycress).

Where: The experiments were conducted at a growth beds at Virginia Tech CRC with Mini T-REX.

Why: To automate plant phenotyping, monitor plant growth, and collect data for 3D reconstruction and analysis. The goal is to develop affordable and readily available automation options for plant research.

How: Using custom-built (Mini T-REX) and commercially available gantry-like robotic systems equipped with cameras and sensors.

Experiment Setting:

Location: The plants were grown in a controlled environment setting.

Influential climatic conditions: Light cycles were controlled, with growth lights mimicking sunlight cycles (16 hours of light). Light intensity could be adjusted based on the crop species.

Processing Methods and Equipment:

Mini T-REX: Intel RealSense L515 LiDAR was used to capture RGB and depth (RGBD) images. SolidWorks and SW2URDF tool were used for CAD design and creating URDF files. ROS2 and MoveIt2 were used for motion planning and control.•

Study Date(s) and Duration:

The data for Mini T-REX was gathered over a span of 30 days in fall 2024

The interval of image capturing was 4 hours every day.

Study Spatial Scale:

Mini T-REX: The gantry footprint is 1650mm x 1650mm. The manipulator has a working radius of 280mm.

Level of True Replication:

The study involved multiple plants (tomatoes).

Sampling Precision:

Mini T-REX: Captured multi-view images from 6 predefined poses for each plant.

Level of Subsampling:

Mini T-REX: RGBD images were captured from 6 different poses.

File Descriptions:

These files include time stamped images captured during the experimental process.