Scholarlink

In this work, the incorporation of a Convolutional Block Attention Module (CBAM) into the DepthAI camera Oak-D Lite for detecting number of apples on trees has been done. The integration procedure encompasses the stages of model building, optimization, conversion to DepthAI format, and deployment on the edge devices. The Convolutional Block Attention Module (CBAM) has been compared with YOLOv3, Faster R-CNN, SSD, and RetinaNet to proved its superioiry. To evaluate the performance we have evaluated five performance parameters that are Precision (.92), Recall (.86), F1-Score (.89), and mAP (.82). Also CBAM model demonstrates a reduced False Positive Rate (FPR) and False Negative Rate (FNR) of 0.0219 and 0.1636 respectively. Moreover, the self created dataset of apple trees has been created to carry out this research. The whole work is tested at shimla region of himacahal pradesh, India.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Fruit Detection Using DepthAI and Convolutional Block Attention Module

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Divyansh Thakur,
  • Vikram Kumar

摘要

In this work, the incorporation of a Convolutional Block Attention Module (CBAM) into the DepthAI camera Oak-D Lite for detecting number of apples on trees has been done. The integration procedure encompasses the stages of model building, optimization, conversion to DepthAI format, and deployment on the edge devices. The Convolutional Block Attention Module (CBAM) has been compared with YOLOv3, Faster R-CNN, SSD, and RetinaNet to proved its superioiry. To evaluate the performance we have evaluated five performance parameters that are Precision (.92), Recall (.86), F1-Score (.89), and mAP (.82). Also CBAM model demonstrates a reduced False Positive Rate (FPR) and False Negative Rate (FNR) of 0.0219 and 0.1636 respectively. Moreover, the self created dataset of apple trees has been created to carry out this research. The whole work is tested at shimla region of himacahal pradesh, India.