Lincoln's Annotated Spatio-Temporal Strawberry Dataset (LAST-Straw)

TL;DR

LAST-Straw delivers a high-resolution spatio-temporal 3D strawberry dataset with rich semantic, instance, and skeleton annotations to enable validation and benchmarking of automated phenotyping pipelines. The paper demonstrates a complete phenotyping workflow, including segmentation, leaf-surface reconstruction, skeletonisation, and tracking, and provides in-silico ground-truth baselines for phenotypes such as leaf area, plant volume, and stem length. By benchmarking multiple reconstruction and skeletonisation methods and introducing a polyline-matching evaluation, the work highlights both the potential and current challenges of real-world 3D plant phenotyping, including occlusion and annotation effort. The dataset and methods aim to accelerate development of next-generation phenotyping tools and enable meaningful cross-study comparisons, with implications for breeding, robotic harvesting, and field-level crop monitoring.

Abstract

Automated phenotyping of plants for breeding and plant studies promises to provide quantitative metrics on plant traits at a previously unattainable observation frequency. Developers of tools for performing high-throughput phenotyping are, however, constrained by the availability of relevant datasets on which to perform validation. To this end, we present a spatio-temporal dataset of 3D point clouds of strawberry plants for two varieties, totalling 84 individual point clouds. We focus on the end use of such tools - the extraction of biologically relevant phenotypes - and demonstrate a phenotyping pipeline on the dataset. This comprises of the steps, including; segmentation, skeletonisation and tracking, and we detail how each stage facilitates the extraction of different phenotypes or provision of data insights. We particularly note that assessment is focused on the validation of phenotypes, extracted from the representations acquired at each step of the pipeline, rather than singularly focusing on assessing the representation itself. Therefore, where possible, we provide \textit{in silico} ground truth baselines for the phenotypes extracted at each step and introduce methodology for the quantitative assessment of skeletonisation and the length trait extracted thereof. This dataset contributes to the corpus of freely available agricultural/horticultural spatio-temporal data for the development of next-generation phenotyping tools, increasing the number of plant varieties available for research in this field and providing a basis for genuine comparison of new phenotyping methodology.

Figures (13)

• Figure 1: The dataset contains two varieties and captures variation in developmental stages, from young to mature plants. Here we provide 2D renders of a sample of 3D meshes for plant B1, as well as all six plants on 07-07-2022 and a top-down view of plant A3 at this timestep. The main volume of the grow bag was manually removed to aid clarity of visualisation for samples of B1. Note - meshes not to scale, best viewed in colour.
• Figure 2: Sample time-series showing a single plant (A2) over the first five time-steps, displaying the colour point clouds (row 1), semantic segmentation (row 2), instance segmentation (row 3) and temporally consistent leaf instance annotations (row 4). Note that the background and scanning table classes have been filtered out.
• Figure 3: Setup of the EinScan Pro 2X Plus operated in 'rapid handheld' mode. Plastic plant and lighting were used for demonstration purposes only.
• Figure 4: Anatomy of a strawberry plant. Left: annotated illustration. Right: annotated sample from the LAST-Straw dataset.
• Figure 5: Ground truth skeletonisation of stem class superimposed on the original point cloud (downsampled for visualisation). Best viewed in colour.