cut-flower disease dataset

Published: 10 October 2021| Version 1 | DOI: 10.17632/nmjd35fcb5.1
Contributor:
Abebe Bekele

Description

This dataset is a dataset that prepared for cut-flower disease detection and classiffication by using computer vision. as it is known cut-flowers are flowers or flower buds that have been cut from the plant bearing it for decorative purpose. it has a great role in balancing a gross domestic product(GDP) of the developing countries like Ethiopia and Kenya. however, it has a big impact on balancing economy, the countries cannot achieve the maximum effort and income due to disease problem. so this dataset is prepared for supporting a researcher those need to support a farmer and investor.the dataset is an image dataset that contain a cut-flowers leaf diseases. black-spot, powdery-mildew, downy-mildew and normal cut-flower leaves are included. in my research that I proposed in Addisababa science and Technology university for partial fulfillment of MSc degree, "developing deep-learning based cut-flower disease detection and classification model " is done by this dataset and achieve a good result in evaluattion matrix. this dataset is prepared through image collecting , image analysis and image labelling procedure. in image collecting(gathering) level the scholar, expert(plant pathologists) and plant managers are participated. a mobile with 64 mpx used for capturing cut-flower leaves image. a distance between a leaves and mobile camera averegically is 30 yards and also in image analysis the blurred image, the image hidden by other leaves, the image captured from a more distance which is not visible is outed during images analysis. finally the images are labelled by using LabelImg tool with their correspondence name. the labelling intension is for multi-label classification which has localization.

Files

Steps to reproduce

this dataset is produced through the following procedure. I) an image data is collected in the field (green house) II) collected images are pre-processed through manually image screening , counting to check balance, reducing unnecessary images III) image labelling(each leaf in a single image are localized and labelled into their correspondence disease category.IV) all labelled image are uploaded onto Roboflow and resizing process take place and the 417*417 size of an images are downloaded and IV) an input images are prepared for feeding into a model.

Institutions

Addis Ababa Science and Technology University

Categories

Deep Learning

License