Determining the effect of color spaces on image based disease classification in agaricus bisporus using squeezenet convolutional neural network

Agaricus bisporus , known as the white-capped mushroom, is among the cultivated mushrooms widely grown worldwide due to its high nutritional value and economic yield. However, this species is threatened by various disease agents including fungal, bacterial, viral and abiotic (non-microbial) agents under intensive production conditions. Early diagnosis of the symptoms caused by these diseases is of great importance in terms of preventing yield losses and minimizing the economic losses associated with them. In this study, it was aimed to classify five different diseases and healthy samples seen in A. bisporus based on images and to evaluate the effects of different color spaces on the classification accuracy. In this context, classification experiments were conducted on image datasets converted into different color space representations, with the SqueezeNet convolutional neural network (CNN) applied separately to each representation. The findings revealed that the XYZ, RGB and YCbCr color space provided higher classification accuracy compared to other formats in the detection of diseases.