Automated detection of grapes and leaves in viticulture with a YOLOv2 neural network

Resumo

In this work, we modeled the problem of detection of fruit and leaves in viticulture for proximal applications as a supervised machine learning task. We created and manually labeled a database of images obtained at Guaspari Winery. In total, the database consists of 11.883 images of bunch of grapes and leaves. We trained a convolutional network with YOLOv2 architecture to locate and classify bunch of grapes and leaves. Quantitative tests have shown results for detection and classification with precision of 100%, recall of 74,22% and F1-Score up to 85,2% for the class “grape”. Also, qualitative tests show that the model generalizes well when tested on photographs of other grape varieties. These results are promising and are moving towards the possibility of application in the field.