Uncovering bias in the PlantVillage dataset

06/09/2022

∙

by   Mehmet Alican Noyan, et al.

∙

54

∙

share

We report our investigation on the use of the popular PlantVillage dataset for training deep learning based plant disease detection models. We trained a machine learning model using only 8 pixels from the PlantVillage image backgrounds. The model achieved 49.0 above the random guessing accuracy of 2.6 PlantVillage dataset contains noise correlated with the labels and deep learning models can easily exploit this bias to make predictions. Possible approaches to alleviate this problem are discussed.