An annotated image dataset of vegetable crops at an early stage of growth for proximal sensing applications

Louis Lac (1) , Barna Keresztes (1) , Marine Louargant (2) , Marc Donias (1) , Jean-Pierre da Costa (1)

1 IMS - Laboratoire de l'intégration, du matériau au système
2 CTIFL - Centre Technique Interprofessionnel des Fruits et Légumes

This article introduces a dataset of 2 801 images of vegetable crops. Maize (Zea mays), bean (Phaseolus vulgaris) and leek (Allium ampeloprasum) crops at an early stage of development (between 2 and 5 weeks from seeding of transplanting) are supported. Two kinds of annotations are provided: (i) bounding boxes enclosing the crops of interest or their stems, weeds being left apart, and (ii) crop structures in the form of star graphs whose vertices are the plant organs (stems and leaves) and whose edges represent the connections between them. The images have been captured in various production and experimentation plots in France using an acquisition module which controls light conditions. They present a wide variety of soil conditions, weed infestation and growth stages. This dataset can benefit precision hoeing and in-field crop monitoring applications that are based on proximal imagery.

Mots clés: Vegetable crops; Crop structures; Precision hoeing; Phenotyping; Leaf counting; Precision agriculture; Crop detection; Deep learning

Domaines: Base de données [cs.DB]; Traitement des images [eess.IV]; Sciences et techniques de l'agriculture; Botanique

Louis Lac, Barna Keresztes, Marine Louargant, Marc Donias, Jean-Pierre da Costa. An annotated image dataset of vegetable crops at an early stage of growth for proximal sensing applications. Data in Brief, 2022, 42, pp.108035. ⟨10.1016/j.dib.2022.108035⟩. ⟨hal-03728222⟩