The computational analysis of phenotypes from images of plants remains an open challenge in computer vision. In order to promote the development of new algorithms in this emerging field, Dr. Sruti Das Choudhury has taken the initiative to release benchmark datasets time to time to promote plant phenotyping research around the world tailored to computer vision and artificial intelligence based research problems in agriculture. The images of the dataets are captured using the LemnaTec Scanalyzer 3D High Throughout Plant Phenotyping facility at the University of Nebraska-Lincoln, UNL, USA.
Released Datasets:
UNL Plant Phenotyping Dataset A: Panicoid Phenomap-1
This dataset consists of images of 31 diverse maize inbreds as well as a number of representative genotypes from other panicoid grain crops. Plants were imaged once per day using the visible light camera. The dataset contains high resolution RGB images taken from the top, front, and side (90 degree rotation) for 176 total plants for 27 days.
The format of the image filename in Dataset A is 'PlantID-GenotypeID.png'.
This dataset is available for download from this link.
If you use this dataset, please cite the following publication:
S. D. Choudhury, V. Stoerger, A. Samal, J. C. Schnable, Z. Liang, J-G. Yu, Automated Vegetative Stage Phenotyping Analysis of Maize Plants using Visible Light Images, KDD workshop on Data Science for Food, Energy and Water (KDD -DSFEW), San Francisco, California, USA, 2016.
UNL Plant Phenotyping Dataset B: Component Plant Phenotyping Dataset (UNL-CPPD)
This dataset consists of images of 13 maize plants for the two views: side view 0 degree and side view 90 degree. Plants were imaged once per day using the visible light camera of the UNL Lemnatec Scanalyzer 3D high throughput phenotyping facility.
The dataset contains original images, groundtruth and annotated images to facilitate research advancements in the field of component phenotyping analysis. This dataset is available for download from this link.
The corresponding publication for citation is given below: S. D. Choudhury, S. Bashyam, Y. Qui, A. Samal, T. Awada, Holistic and Component Plant Phenotyping using Visible Light Image Sequence, Plant Methods, 14:35, 2018.
UNL 3D Plant Phenotyping Dataset: (UNL-3DPPD)
This dataset consists of images of 20 maize plants & 20 sorghum plants for 10 side views. Plants were imaged once per day using the visible light camera of the UNL Lemnatec Scanalyzer 3D high throughput phenotyping facility.
The dataset is introduced to facilitate research advancements in the field of 3D plant phenotyping analysis. This dataset is available for download from this link.
The checkerboard images for camera calibration can be downloaded from here.
The corresponding publication for citation is: S. D. Choudhury, S. Maturu, V. Stoerger, A. Samal, T. Awada, Leveraging Image Analysis to Compute 3D Plant Phenotypes based on Voxel-Grid Plant Reconstruction, Frontiers in Plant Science, 11(521431), 2020.
FlowerPheno Dataset:
This dataset consists of images of Coleus, Canna and Sunflower plants captured from 10 side views. This dataset is intended to promote research in flower phenotyping analysis using deep neural networks.
This dataset is available for download from this link.
The corresponding publication for citation is: S. D. Choudhury, S. Guha, A. Das, A. K. Das, A. Samal, T. Awada, FlowerPhenoNet: A Deep Learning System for Flower Detection using Image Sequences for Temporal Phenotyping, Plant Phenomics, (under review), 2021.
Cosegmentation for Plant Phenotyping (CosegPP) Data Repository
The dataset is introduced to facilitate research advancements in the field of multimodal cosegmentation in plant phenotyping analysis. This dataset is available for download from this link.
UNL- Maize Emergence Dataset (UNL-MED)
The emergence timing of a plant, i.e., the time at which the plant is first visible from the surface of the soil, is an important phenotypic event as an indicator of successful establishment and growth of a plant. The dataset contains top-view time-lapse images of maize plants starting before the occurrence of the emergence event and continues until maize seedlings are about one inch tall and also binary masks. This dataset is introduced to facilitate development of algorithms for coleoptile emergence timing detection and tracks its growth from a time-lapse sequence with cluttered background and extreme variations in illumination. This dataset is available for download from this link.
CottonHusker Dataset: Cotton Detection For Robotic Cotton-Picking
The dataset was created for cotton detection in simulated cotton field. This dataset is available for download.