Data from: Apple flower detection using deep convolutional networks

To optimize fruit production, a portion of the flowers and fruitlets of apple trees must be removed early in the growing season. The proportion to be removed is determined by the bloom intensity, i.e., the number of flowers present in the orchard. Several automated computer vision systems have been proposed to estimate bloom intensity, but their overall performance is still far from satisfactory even in relatively controlled environments. With the goal of devising a technique for flower identification which is robust to clutter and to changes in illumination, this paper presents a method in which a pre-trained convolutional neural network (CNN) is fine-tuned to become specially sensitive to flowers. Experimental results on a challenging dataset demonstrate that our method significantly outperforms three approaches that represent the state of the art in flower detection, with recall and precision rates higher than 90%. Moreover, a performance assessment on three additional datasets previously unseen by the network, which consist of different flower species and were acquired under different conditions, reveals that the proposed method highly surpasses baseline approaches in terms of generalization capability. This dataset comprises mp4 video sequences illustrating each combination of datasets and methods. Resources in this dataset:Resource Title: Supplementary data - Video mmc1 (7MB). File Name: 1-s2.0-S016636151730502X-mmc1.mp4Resource Description: Dataset = AppleA. Method on left-hand side: second baseline algorithm mentioned in the paper, where HSV is hue-saturation-value, and 'Bh' is Bhattacharyya distance. Method on right-hand side: our proposed method, the CNN + SVM, where CNN = convolutional neural network and SVM = support vector machine. True Positives (blue), False Positives (cyan), and False Negatives (red).Resource Title: Supplementary data - Video mmc2 (7MB). File Name: 1-s2.0-S016636151730502X-mmc2.mp4Resource Description: Dataset = AppleA. Method on left-hand side: third baseline algorithm mentioned in the paper, HSV + SVM, where HSV is hue-saturation-value and SVM is support vector machine. Method on right-hand side: our proposed method, the CNN + SVM, where CNN = convolutional neural network and SVM = support vector machine. True Positives (blue), False Positives (cyan), and False Negatives (red).Resource Title: Supplementary data - Video mmc3 (7MB). File Name: 1-s2.0-S016636151730502X-mmc3.mp4Resource Description: Dataset = AppleA. Method on left-hand side: first baseline algorithm mentioned in the paper, where HSV is hue-saturation-value.
Method on right-hand side: our proposed method, the CNN + SVM, where CNN = convolutional neural network and SVM = support vector machine. True Positives (blue), False Positives (cyan), and False Negatives (red).Resource Title: Supplementary data - Video mmc4 (3MB). File Name: 1-s2.0-S016636151730502X-mmc4.mp4Resource Description: Dataset = AppleB. Method on left-hand side: third baseline algorithm mentioned in the paper, HSV + SVM, where HSV is hue-saturation-value and SVM is support vector machine. Method on right-hand side: our proposed method, the CNN + SVM, where CNN = convolutional neural network and SVM = support vector machine. True Positives (blue), False Positives (cyan), and False Negatives (red).Resource Title: Supplementary data - Video mmc5 (3MB). File Name: 1-s2.0-S016636151730502X-mmc5.mp4Resource Description: Dataset = AppleC. Method on left-hand side: third baseline algorithm mentioned in the paper, HSV + SVM, where HSV is hue-saturation-value and SVM is support vector machine. Method on right-hand side: our proposed method, the CNN + SVM, where CNN = convolutional neural network and SVM = support vector machine. True Positives (blue), False Positives (cyan), and False Negatives (red).Resource Title: Supplementary data - Video mmc6 (3MB). File Name: 1-s2.0-S016636151730502X-mmc6.mp4Resource Description: Dataset = Peach. Method on left-hand side: third baseline algorithm mentioned in the paper, HSV + SVM, where HSV is hue-saturation-value and SVM is support vector machine. Method on right-hand side: our proposed method, the CNN + SVM, where CNN = convolutional neural network and SVM = support vector machine. True Positives (blue), False Positives (cyan), and False Negatives (red).

Data and Resources

Field Value
accessLevel public
bureauCode {005:18}
catalog_@context https://project-open-data.cio.gov/v1.1/schema/catalog.jsonld
catalog_conformsTo https://project-open-data.cio.gov/v1.1/schema
catalog_describedBy https://project-open-data.cio.gov/v1.1/schema/catalog.json
identifier 10.15482/USDA.ADC/1503382
license https://www.usa.gov/publicdomain/label/1.0/
modified 2024-02-15
programCode {005:040}
publisher Agricultural Research Service
resource-type Dataset
source_datajson_identifier true
source_hash ead4d4aef4db01d9bb977d8804c2e2a81aac71a44c3290e6748eacf3131896a6
source_schema_version 1.1
Groups
  • AmeriGEOSS
  • National Provider
  • North America
Tags
  • AmeriGEO
  • AmeriGEOSS
  • CKAN
  • GEO
  • GEOSS
  • National
  • North America
  • United States
  • apple-flower-detection
  • ars
  • bloom-intensity-estimation
  • convolutional-neural-networks
  • data-gov
  • deep-learning
  • orchard-automation
isopen False
license_id us-pd
license_title us-pd
maintainer Tabb, Amy
maintainer_email amy.tabb@ars.usda.gov
metadata_created 2025-09-23T19:10:59.239909
metadata_modified 2025-09-23T19:10:59.239914
notes <p>To optimize fruit production, a portion of the flowers and fruitlets of apple trees must be removed early in the growing season. The proportion to be removed is determined by the bloom intensity, i.e., the number of flowers present in the orchard. Several automated computer vision systems have been proposed to estimate bloom intensity, but their overall performance is still far from satisfactory even in relatively controlled environments. With the goal of devising a technique for flower identification which is robust to clutter and to changes in illumination, this paper presents a method in which a pre-trained convolutional neural network (CNN) is fine-tuned to become specially sensitive to flowers. Experimental results on a challenging dataset demonstrate that our method significantly outperforms three approaches that represent the state of the art in flower detection, with recall and precision rates higher than 90%. Moreover, a performance assessment on three additional datasets previously unseen by the network, which consist of different flower species and were acquired under different conditions, reveals that the proposed method highly surpasses baseline approaches in terms of generalization capability. This dataset comprises mp4 video sequences illustrating each combination of datasets and methods. </p><div><br>Resources in this dataset:</div><br><ul><li><p>Resource Title: Supplementary data - Video mmc1 (7MB).</p> <p>File Name: 1-s2.0-S016636151730502X-mmc1.mp4</p><p>Resource Description: Dataset = AppleA. </p> <p>Method on left-hand side: second baseline algorithm mentioned in the paper, where HSV is hue-saturation-value, and 'Bh' is Bhattacharyya distance. </p> <p>Method on right-hand side: our proposed method, the CNN + SVM, where CNN = convolutional neural network and SVM = support vector machine.</p> <p>True Positives (blue), False Positives (cyan), and False Negatives (red).</p></li><br><li><p>Resource Title: Supplementary data - Video mmc2 (7MB).</p> <p>File Name: 1-s2.0-S016636151730502X-mmc2.mp4</p><p>Resource Description: Dataset = AppleA. </p> <p>Method on left-hand side: third baseline algorithm mentioned in the paper, HSV + SVM, where HSV is hue-saturation-value and SVM is support vector machine.</p> <p>Method on right-hand side: our proposed method, the CNN + SVM, where CNN = convolutional neural network and SVM = support vector machine.</p> <p>True Positives (blue), False Positives (cyan), and False Negatives (red).</p></li><br><li><p>Resource Title: Supplementary data - Video mmc3 (7MB).</p> <p>File Name: 1-s2.0-S016636151730502X-mmc3.mp4</p><p>Resource Description: Dataset = AppleA. </p> <p>Method on left-hand side: first baseline algorithm mentioned in the paper, where HSV is hue-saturation-value. </p> <p>Method on right-hand side: our proposed method, the CNN + SVM, where CNN = convolutional neural network and SVM = support vector machine.</p> <p>True Positives (blue), False Positives (cyan), and False Negatives (red).</p></li><br><li><p>Resource Title: Supplementary data - Video mmc4 (3MB).</p> <p>File Name: 1-s2.0-S016636151730502X-mmc4.mp4</p><p>Resource Description: Dataset = AppleB.</p> <p>Method on left-hand side: third baseline algorithm mentioned in the paper, HSV + SVM, where HSV is hue-saturation-value and SVM is support vector machine.</p> <p>Method on right-hand side: our proposed method, the CNN + SVM, where CNN = convolutional neural network and SVM = support vector machine.</p> <p>True Positives (blue), False Positives (cyan), and False Negatives (red).</p></li><br><li><p>Resource Title: Supplementary data - Video mmc5 (3MB).</p> <p>File Name: 1-s2.0-S016636151730502X-mmc5.mp4</p><p>Resource Description: Dataset = AppleC.</p> <p>Method on left-hand side: third baseline algorithm mentioned in the paper, HSV + SVM, where HSV is hue-saturation-value and SVM is support vector machine.</p> <p>Method on right-hand side: our proposed method, the CNN + SVM, where CNN = convolutional neural network and SVM = support vector machine.</p> <p>True Positives (blue), False Positives (cyan), and False Negatives (red).</p></li><br><li><p>Resource Title: Supplementary data - Video mmc6 (3MB).</p> <p>File Name: 1-s2.0-S016636151730502X-mmc6.mp4</p><p>Resource Description: Dataset = Peach.</p> <p>Method on left-hand side: third baseline algorithm mentioned in the paper, HSV + SVM, where HSV is hue-saturation-value and SVM is support vector machine.</p> <p>Method on right-hand side: our proposed method, the CNN + SVM, where CNN = convolutional neural network and SVM = support vector machine.</p> <p>True Positives (blue), False Positives (cyan), and False Negatives (red).</p></li></ul><p></p>
num_resources 6
num_tags 15
title Data from: Apple flower detection using deep convolutional networks