Skip to Main content Skip to Navigation
Journal articles

Novel data augmentation strategies to boost supervised segmentation of plant disease

Clément Douarre 1 Carlos Crispim-Junior 1 Anthony Gelibert 2 Laure Tougne 1 David Rousseau 3
1 imagine - Extraction de Caractéristiques et Identification
LIRIS - Laboratoire d'InfoRmatique en Image et Systèmes d'information
2 CTSYS - Conception et Test de SYStèmes embarqués
LCIS - Laboratoire de Conception et d'Intégration des Systèmes
3 Images et Modèles
CREATIS - Centre de Recherche en Acquisition et Traitement de l'Image pour la Santé
Abstract : Annotation of images in supervised learning is notably costly and time-consuming. In order to reduce this cost, our objective was to generate images from a smalldataset of annotated images, and then use those synthesized images to help the network’s training process. In this article, we tackled for illustration with agriculturalmaterial the difficult segmentation task of apple scab on images of apple plant canopy by using convolutional neural networks. We devised two novel methods ofgenerating data for this use case: one based on a plant canopy simulation and the other on Generative Adversatial Networks (GANs). As a result, we found thatsimulated data could provide an important increase in segmentation performance, up to a 17% increase of F1 score (a measure taking into account precision andrecall), compared to segmenting with weights initialized on ImageNet. In this way, we managed to obtain, with small datasets, higher segmentation scores than theones obtained with bigger datasets if using no such augmentations. Moreover, we left our annotated dataset of scab available for the plant science imaging com-munity. The proposed method is of large applicability for plant diseases observed at a canopy scale.
Document type :
Journal articles
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-02330900
Contributor : Douarre Clément <>
Submitted on : Thursday, October 24, 2019 - 11:00:23 AM
Last modification on : Thursday, November 19, 2020 - 1:01:31 PM

Identifiers

Citation

Clément Douarre, Carlos Crispim-Junior, Anthony Gelibert, Laure Tougne, David Rousseau. Novel data augmentation strategies to boost supervised segmentation of plant disease. Computers and Electronics in Agriculture, Elsevier, 2019, 165, pp.104967. ⟨10.1016/j.compag.2019.104967⟩. ⟨hal-02330900⟩

Share

Metrics

Record views

178