A comprehensive dataset of mandarin leaf images for classification

Abstract

The research is devoted to mandarin leaf classification using a deep learning approach. Citrus cultivation has a big market in Bangladesh. In citrus, mandarin is one of them. For the research, we collected images from the Natore district in Bangladesh. The datasets are classifying with the Mandarin varieties are cultivated in Bangladesh. Image classification has been done with Mandarin veritas, such as China Mishti, Nagpuri, Darjeeling, and Meandering. The images were collected in 2608 × 4624 pixels from Realme 8. The raw images are in four varieties, and the total image count is 1917. The image is also attached with augmented images into different augmented sub-folders with a count of eight thousand images. The augmented images are included with image rotations and contrast enhancement. The augmented image count is 8000. All the collected and augmented images are included with healthy images, which help to classify mandarin leaves using the deep-learning approach with reasonable accuracy. The datasets are open for the researchers for their innovative research to advance the leaf classification technique and contribute their research and development to the agriculture domain for enhancing the classification technique with an automated system. The dataset comprises only healthy leaf images, which not only aids in the accurate classification of mandarin leaves using deep learning techniques but also supports early identification of healthy foliage. This is crucial for gardeners and farmers, as it can help detect issues such as mis-grafting in trees, enabling timely interventions.