Scholarlink

<p>This article presents a dataset of oil palm Fresh Fruit Bunches (FFBs) images from commercial plantations in Central Kalimantan, Indonesia, focusing on five maturity stages: Unripe, Underripe, Ripe, Flower, and Abnormal. The data collection involved smartphone video recordings of unharvested trees from multiple angles under varying conditions. Video frames were extracted and expertly annotated using Computer Vision Annotation Tool (CVAT), with annotations exported in Common Objects in Context (COCO) format suitable for object detection tasks. It has 10,207 images in its training set, 2,896 in the validation set, and 1,400 in the test set, which are supplemented using data augmentation to handle class imbalance and increase variation. These images have real-world complications arising from partial visibility, low contrast, occlusion, and blurriness. It forms the basis that will support the development of deep learning models for detection and classification of FFB, particularly for monitoring of harvest times, yield prediction, and optimization of resources in plantation operations.</p>

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

Oil Palm Fruits Dataset in Plantations for Harvest Estimation Using Digital Census and Smartphone

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年6月10日

查看全文
  • Suharjito,
  • Martinus Grady Naftali,
  • Gregory Hugo,
  • Muhammad Reza Azhar Priyadi,
  • Muhammad Asrol,
  • Ditdit Nugeraha Utama

摘要

This article presents a dataset of oil palm Fresh Fruit Bunches (FFBs) images from commercial plantations in Central Kalimantan, Indonesia, focusing on five maturity stages: Unripe, Underripe, Ripe, Flower, and Abnormal. The data collection involved smartphone video recordings of unharvested trees from multiple angles under varying conditions. Video frames were extracted and expertly annotated using Computer Vision Annotation Tool (CVAT), with annotations exported in Common Objects in Context (COCO) format suitable for object detection tasks. It has 10,207 images in its training set, 2,896 in the validation set, and 1,400 in the test set, which are supplemented using data augmentation to handle class imbalance and increase variation. These images have real-world complications arising from partial visibility, low contrast, occlusion, and blurriness. It forms the basis that will support the development of deep learning models for detection and classification of FFB, particularly for monitoring of harvest times, yield prediction, and optimization of resources in plantation operations.