DeepGrapes: Precise Detection of Grapes in Low-resolution Images

摘要

We proposed a deep convolutional network aimed at detection of single grapes of white varieties in low-resolution colour images. With accuracy, precision and recall at 97.35 %, 96.53 %, and 98.27 %, the proposed grape detector is a state of the art solution. The detector is intended for practical applications that require high exactitude, such as in the process of yield estimation.