A fully labelled image dataset of banana leaves deficient in nutrients

Agriculture is the art and science of cultivating the soil, growing crops and raising livestock. Enhancing crop yield becomes essential to improve economy in agriculture sector. Crops need 16 essential nutrients in balanced factor for their proper growth. Deficiency among these essential nutrients causes stunted growth often leading to significant crop loss. Symptoms of nutrient deficiency in plants can be observed visually which needs to be diagnosed correctly to rectify the problem, so that plants can grow healthily, increasing its yield. Among various crops, banana is one of the staple foods for millions of people across country and world. They contain essential nutrients that can have a defending impact on health of the human beings. The year-round availability, affordability, varietal range, taste, nutritive of banana and medicinal value makes it the favorite fruit among all classes of people. In addition, it also has good export potential. Few of the symptoms due to nutrient's deficiency in banana leaves are like curling of leaves, appearance of yellow strips, yellowing of the leaves, bluish color of young leaves. Deficiency symptoms can be visualized prominently on the leaves of the plant. Further, Machine Learning models can be developed to detect nutrient deficiency in leaves and help farmers in taking relevant measures. Thus, a fully labelled dataset becomes essential to train and test these models to detect nutrient deficiency accurately. The dataset created consists of banana leaf images of various categories like Musa acuminata (Dwarf Cavendish), Robusta, Rasthali, Poovan, Monthan, Elakkibale. Images depict deficiency in eight class of nutrients: boron, calcium, iron, potassium, manganese, magnesium, sulphur and zinc. Table 1 summarizes the essential nutrients and their deficiency symptoms visible on the leaves. Dataset also contains images of healthy leaves. Machine Learning Models can be developed by researchers and students and train them by using the created banana dataset to obtain high accuracy.


a b s t r a c t
Agriculture is the art and science of cultivating the soil, growing crops and raising livestock. Enhancing crop yield becomes essential to improve economy in agriculture sector. Crops need 16 essential nutrients in balanced factor for their proper growth. Deficiency among these essential nutrients causes stunted growth often leading to significant crop loss. Symptoms of nutrient deficiency in plants can be observed visually which needs to be diagnosed correctly to rectify the problem, so that plants can grow healthily, increasing its yield. Among various crops, banana is one of the staple foods for millions of people across country and world. They contain essential nutrients that can have a defending impact on health of the human beings. The year-round availability, affordability, varietal range, taste, nutritive of banana and medicinal value makes it the favorite fruit among all classes of people. In addition, it also has good export potential. Few of the symptoms due to nutrient's deficiency in banana leaves are like curling of leaves, appearance of yellow strips, yellowing of the leaves, bluish color of young leaves. Deficiency symptoms can be visualized prominently on the leaves of the plant. Further, Machine Learning models can be developed to detect nutrient deficiency in leaves and help farmers in taking relevant measures. Thus, a fully labelled dataset becomes essential to train and test these models to detect nutrient deficiency accurately. The dataset created consists of banana leaf images of various categories like Musa acuminata (Dwarf Cavendish), Robusta, Rasthali, Poovan, Monthan, Elakkibale. Images depict deficiency in eight class of nutrients: boron, calcium, iron, potassium, manganese, magnesium, sulphur and zinc. Table  1 summarizes the essential nutrients and their deficiency symptoms visible on the leaves. Dataset also contains images of healthy leaves. Machine Learning Models can be developed by researchers and students and train them by using the created banana dataset to obtain high accuracy.

Value of the Data
Artificial Intelligence has taken over every field in the world -Marketing, Agriculture, Industry, Finance, Technology, etc. Machine Learning and Deep Learning models need to be trained with huge amount of data to obtain high precision results [1,2] . Thus, these models are data hungry.
• The dataset of nutrient deficient banana leaves contains images belonging to eight class of nutrients. • The dataset provides raw data of 30 0 0 images and is augmented to 70 0 0 + images of eight classes. • The images are manually labelled by an agriculture scientist.
• The present comprehensive dataset can help researchers / students to evaluate their Machine / Deep Learning models and can achieve high precision results.

Objective
• Create image dataset of banana leaves which are deficient in nutrients.
• Collect images of various category of banana plants across different locations.
• Gather diverse data by capturing images of the leaves from various angles, different perspectives and in variable lighting conditions.

Data Description
Banana is a rich source of carbohydrate and is rich in vitamins particularly vitamin B. The fruit is easy to digest, free from fat and cholesterol. It helps in reducing risk of heart diseases when used regularly and is recommended for patients suffering from high blood pressure, arthritis, ulcer, gastroenteritis and kidney disorders. Banana is considered a foodie or heavy feeder plant, so soil fertility is really important for it [3] .
Banana is a quick growing and short-lived plant. It responds very well to the application of balanced nutrients. Deficiency of the essential nutrients leads to stunted growth of the plants thus decreasing their yield. Symptoms of deficiency in nutrients can be prominently seen on the leaves of the plants [4,5] . Machine Learning models can be designed to detect the deficiency symptoms by processing the images of the leaves. Hence, dataset of leaf images which are deficient in nutrients become essential to train Machine learning models.
The dataset created consists of images of banana leaves which are deficient in nutrients. Images captured belong to various categories of banana across different locations in Karnataka, India. The dataset contains eight folders labelled with each nutrient class namely: Boron, Calcium, Iron, Potassium, Magnesium, Manganese, Sulphur and Zinc. Table 2 tabulates the sample of images of each nutrient category with their deficiency symptoms [5] . In addition, healthy leaves are stored in another folder. Leaves can also be found to be deficient in multiple nutrients [6] . Hence, more than one folder can contain same leaf images. The captured raw images are processed by resizing it to a standard size of 225 × 225. Images are cropped manually to remove unwanted parts and background is set to black so that designed Machine learning models get trained with uniform images and report high accuracy [2] . In addition, the accuracy of the Machine/ Deep Learning model depends on the richness of the dataset [6,7] . Hence, the created dataset containing 30 0 0 raw images, is augmented to 70 0 0 + images. Augmentation is carried out in such a manner that each category has approximately same number of images, so that Machine /Deep Learning models should not undergo training with imbalanced dataset.
(see Table 1 ) Magnesium Yellow discolouration is observed in the mid blade and midrib portion, the margins of the leaf remain green. Purple mottling of the petioles, Leaf symptoms include marginal yellowing 4.
Sulphur Yellow or white appearance of young leaves, necrotic patches on the leaf margins, leaves take on a silver-green colouring 5.
Iron Younger leaves become pale green or white in colour 6.
Zinc Anthocyanin pigmentation appear on its underside, yellow to white stripes between the secondary veins, alternating yellow and green stripes 7.
Manganese Narrow green edge appears at the leaf margins of second or third youngest leaf, which further spreads along the main veins towards the midrib 8.
Calcium Youngest leaf become abnormal in size, leaf margins become dry and veins become erupted and cause the spike leaf

Design and Materials
Mobile camera is used to capture banana leaf images to create dataset of nutrient deficiency. Mobile cameras of Samsung M20, A12, Motorola G5, ONEPlus, RedMi, were used to capture images. Images are captured in different environmental conditions like in the morning and evening. Few images were also taken in the afternoon due to which images have shaded regions. Certain Images also contain dew drops as they were taken in the morning time. Images are acquired from various locations in Karnataka, India. Table 2 The sample banana leaf images which are deficient in nutrients of 8 categories.   Table 3 Specifications of the camera and the Images captured. Table 6 Description and count of images belonging to eight class of nutrients.