Deep Learning for Grape Cluster Segmentation

Authors

  • Abhiram Patil
  • Parikshit Mahalle

DOI:

https://doi.org/10.37591/rtpl.v8i2.123

Keywords:

YOLO v3, mask RCNN, U-net, modified U-net, performance, segmentation of grape cluster

Abstract

Grape cluster identification and its segmentation for the purpose of vineyard total weight prediction tasks indicate the need for more accurate segmentation atomization. The Grape Cluster Segmentation challenge is supplied as an answer the usage of deep neural community fashions including YOLO v3, Mask RCNN, and U-net. In the sense of a modified U-net model for segmenting grapes using training and testing strategies, this contribution contributes to the validation of results. The results were obtained for the accuracy of classifying pixels as part of a cluster of grapes or outside the cluster, and the comparison results show improved segmentation with the modified U-net. Accuracy, accuracy, and memory analysis are performed, and the comparison model shows better results.

Published

2022-01-18