Sorghum Plant Centers 2016 Dataset

The PhenoSorg Project

This dataset consists of a total of 60,000 images of sorghum plants taken from an Unmaned Aerial Vehicle. There is a total of 15,208 unique plants with their center labeled.

Summary

This dataset was generated as follows. On June 21, 2016, a UAV was flown at an altitude of 40 meters over a crop of sorghum plants located at Purdue University. The images were stitched together to generate a 6,000 x 12,000 orthomosaic of 0.75 cm/pixel resolution. After labeling all plants centers, the orthomosaic area was split as 80-10-10 % to generate the training, validation and testing subsets. These areas are highlighted in red, green, and blue in the image above. Within each region, random image crops were extracted (50,000 for training, 5,000 for validation, and 5,000 for testing). The height and width of the random crops are uniformly distributed between 100 and 600 pixels. Note that some of these image crops may overlap, but they are restricted to their area. For more details on how this dataset was generated, please see the publications below.

Download

Download the dataset here. The same dataset with the images resized to 256x256 can be downloaded from here. The plant centers are modified appropriately.

Citations

If you use this dataset, please cite the following publications:

Description

Each directory contains a CSV file called gt.csv, which contains the groundtruth as follows.

The CSV file describes the groundtruth of all the images in the same directory.

Each row contains the count (number) and location of all plants in the image with a particular filename. The plant locations are in (y, x) format, being the origin the most top left pixel, y being the pixel row number, and x being the pixel column number.

Principal Investigator

Edward J. Delp (Principal Investigator), The Charles William Harrison Distinguished Professor of Electrical and Computer Engineering and Professor of Biomedical Engineering, Purdue University

Students

Javier Ribera (PhD Student) Electrical and Computer Engineering, Purdue University

Yuhao Chen (PhD Student) Electrical and Computer Engineering, Purdue University

This work is supported by the Advanced Research Projects Agency-Energy (ARPA-E) as part of the Automated Sorghum Phenotyping and Trait Development Platform program.

The information, data, or work presented herein was funded in part by the Advanced Research Projects Agency-Energy (ARPA-E), U.S. Department of Energy, under Award Number DE-AR0000593. The views and opinions of the authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.