A key resource for scientific research

The large-scale repository of Arabidopsis thaliana RGB images with rosette and leaf annotations

Partners

La Trobe Institute for Sustainable Agriculture & Food (LISAF)

About datasets

Arabidopsis thaliana Phenotyping Datasets

This webpage provides large-scale RGB imaging datasets of various Arabidopsis thaliana ecotypes. Three high-throughput plant phenotyping experiments were conducted in a fully controlled growth chamber. The datasets include raw RGB images, manually labelled plant canopies, and individual leaf masks to facilitate open-source phenotyping data sharing.

These resources support applications such as plant trait extraction, statistical analysis, and deep learning model development for computer vision tasks such as plant segmentation, stress detection, and growth prediction. By enabling research in precision agriculture, plant development, and AI-driven phenotyping, the datasets contribute to advancing plant science and image-based plant phenotyping.

Raw RGB Images

Plant Labels

Leaf Labels

Growth Sequence

Figure 1: Experimental setup for data collection using an X-Y indoor phenotyping system — Figure 1. Experimental setup for data collection using an X-Y indoor phenotyping system

Experiments conducted

Phenotyping Experiments

Three phenotyping experiments were conducted at La Trobe University, Australia, to extensively study Arabidopsis thaliana growth under controlled conditions, as shown in Figure 1.

The first experiment included 38 ecotypes, imaged twice daily. The second experiment focused on eight selected ecotypes, with increased imaging frequency. Both experiments used a high-throughput system with an adjustable-height table and automated top-view RGB imaging.

The third experiment, conducted earlier, utilized 30 fixed stereo cameras to capture plant growth from multiple angles over time. Environmental parameters, including light cycles and temperature, were carefully regulated across all experiments.

Data preparation involved image cropping and organization to ensure high-quality datasets for further analysis.

Plant and Leaf Datasets

Data Organization and Structure

The datasets include plant-level and leaf-level data collected from Arabidopsis experiments. The plant datasets (plant_ds1, plant_ds2, and plant_ds3) contain raw RGB images, and the first two also include manually labelled plant masks, as shown in Figure 2A and Figure 2C. Plant datasets captured whole plant growth over time, with structured folders for different ecotypes and replicates.

The leaf datasets (PhenoLeaf-TS, leaf_dataset1, and leaf_dataset2) provide individual leaves labelled using different colours as well as leaf binary masks, as shown in Figure 2B and Figure 2D. These datasets focused on individual leaves, tracking both visible and occluded leaves through binary masks. The coloured leaf labels available in PhenoLeaf-TS ensure consistent leaf identification throughout the plant's life cycle. All datasets follow a hierarchical structure, allowing organized storage and retrieval of plant and leaf data.

A

Figure 2A: Examples of raw RGB images of the Col-0 Arabidopsis ecotype with corresponding rosette labels.

B

Figure 2B: Examples of raw RGB images of the Ler-1 Arabidopsis ecotype with corresponding coloured leaf labels.

C

Figure 2C: Examples of raw RGB images of the Ler-1 Arabidopsis ecotype with segmented images.

D

Figure 2D: Examples of raw RGB images of the Uk-1 Arabidopsis ecotype with corresponding coloured leaf labels.

Figure 2. (A) Examples of raw RGB images of the Col-0 Arabidopsis ecotype with corresponding rosette labels; (B) Examples of raw RGB images of the Ler-1 Arabidopsis ecotype with corresponding coloured leaf labels (Sarić et al., 2026); (C) Examples of raw RGB images of the Ler-1 Arabidopsis ecotype with segmented images; (D) Examples of raw RGB images of the Uk-1 Arabidopsis ecotype with corresponding coloured leaf labels (Sarić et al., 2026)

View plant dataset View leaf dataset