LeaData: a novel reference data of digital microscopic leather images for automatic species identification
Data files
Aug 28, 2024 version files 3.30 GB
-
LeaData-V1.zip
470.33 MB
-
LeaData-V2.zip
2.83 GB
-
README.md
5.22 KB
Aug 28, 2024 version files 3.30 GB
-
LeaData-V1.zip
470.33 MB
-
LeaData-V2.zip
2.83 GB
-
README.md
5.21 KB
Abstract
In the leather industry, the mammalian skins of buffalo, cow, goat, and sheep are the permissible materials for leather-making. They serve the trade of quality leather products; hence, the knowledge of animal species in leather is inevitable. The traditional identification techniques are prone to ambiguous predictions due to insufficient reference studies. Indeed, leather image analysis with big data can pave the way for automatic and objective analysis with accurate prediction. Therefore, a novel and unique leather image data, LeaData, is created to facilitate automatic species identification in leather from grain surface analysis. A simple, cheaper, handheld digital microscope with the magnifying parameter 47x is chosen to capture the species-unique grain patterns distributed over the leather surface. Currently, the novel LeaData encloses 38,172 leather images of four species from 137 leather samples. This big data spans leather images with theoretically ideal and practically non-ideal characteristics of grain patterns. It also includes images of grain patterns varying over different body parts. Thus, the LeaData is an adequately larger pool of leather images with diverse behavior. It is also divided into three versions, LeaData-V1, LeaData-V2, and LeaData-V3, based on leather image characteristics, such as ideal, non-ideal, and body part-specific. Presenting LeaData and its versions to the research community of digital image processing and computer vision can help establish a smart leather species identification technique that can be easily accessible by leather specialists, customs officials, and leather product manufacturers. This digitized source of permissible leather species helps enable digitization in leather technology for species identification. In turn, in maintaining biodiversity preservation and consumer protection.
https://doi.org/10.5061/dryad.12jm63z6s
Anjli Varghese, Malathy Jawahar, A. Amalin Prince
Description of the data and file structure
LeaData is a unique image data introduced to digital image processing (DIP) and computer vision fields to establish digitalization in leather technology. It is created to develop a smart identification technique that can quickly determine the animal species in leather without human intervention. This data thereby has a great impact on assisting the leather specialists, customs officials, and leather product manufacturers for quick and easy decision-making. Moreover, the outcome of this data can take part in maintaining biodiversity preservation and consumer protection.
The LeaData spans images of the leather surface background with species-distinct hair pore patterns. It captures the hair-pore behavior of the four most commonly used leather species: buffalo, cow, goat, and sheep. Although these patterns undergo vivid characteristics between species, different leather samples belong to different regions of location, breeds, ages, genders, different parts of the body, etc. Hence, for better species distinction, LeaData encloses leather images with diverse pore characteristics and groups them into different versions.
- LeaData-V1: A small-scale version of LeaData with 1200 leather images of four species exhibiting ideal characteristics. The theoretical features of leather samples learned and surveyed over generations are termed as the ideal characteristics.
- LeaData-V2: A comparatively large-scale version of LeaData with 7600 leather images of four species exhibiting non-ideal characteristics. The leather images with pore features different from the ideal behavior are termed as the non-ideal characteristics. It is practically challenging to determine the animal species accurately from the leather images with non-ideal behavior.
- LeaData-V3: A very large-scale version of LeaData with 10000 leather images with body part-specific characteristics. The pore features ideally posses varied distribution among different parts of the body which are called as the body part-specific characteristics. Learning from diverse pore behavior can enhance the image understanding for reliable decision-making.
Currently, the images in LeaData-V1 and LeaData-V2 are uploaded, meanwhile the images in LeaData-V3 will be uploaded in future.
The images in LeaData-V1 and LeaData-V2 can be used for research purposes and the authors must cite the database and related papers in their articles. Meanwhile, a written permission is required for any modifications or commercial work.
Files and variables
File: LeaData-V1.zip
Description: It is a small-scale data of leather images with ideal behavior. It consists of 1200 leather images grouped into four species. The main folder has four sub-folders: Buffalo, Cow, Goat, and Sheep, and each sub-folder has 300 leather images. The filename of each image in every folder starts with the first letter of the folder describing its species, followed by its body part, sample number, and image number. Also, the images are saved in JPG format with a .jpg extension.
For example, in the filename BBTS1 (1).jpg:
- B indicates the species Buffalo
- BT indicates the body part BuTt
- S1 indicates the leather Sample number 1
- (1) indicates image number 1 captured from sample S1
- .jpg indicates the JPG format in which the image is saved
Researchers can use LeaData-V1 to apply different digital image processing and machine learning techniques to mathematically learn and quantify the ideal features of species' images.
File: LeaData-V2.zip
Description: It is a large-scale data of leather images with non-ideal behavior. It consists of 7600 leather images grouped into four species. The main folder has four sub-folders: Buffalo, Cow, Goat, and Sheep, and each sub-folder has 1900 leather images. The filename of each image in every folder starts with the first letter of the folder describing its species, followed by its body part, sample number, and image number. Also, the images are saved in JPG format with a .jpg extension.
Researchers can use LeaData-V2 to apply different digital image processing and machine learning techniques to mathematically learn and quantify the ideal features of species' images. Moreover, it is suitable to explore deep learning techniques to develop an automatic species identification model. Indeed, combining and learning from LeaData-V1 and LeaData-V2 can further enhance the species-specific pore description with diverse behavior for efficient species prediction.
The leather image acquisition process is executed and assisted by the concerned leather experts of the Central Leather Research Institute (CLRI), Chennai, India. They provided 137 leather samples of four animal species, buffalo, cow, goat, and sheep, with diverse behaviors for acquisition. Celestron handheld digital microscope pro is used to capture species-specific grain patterns. The acquisition process is initiated with 47x magnification and 1024 x 1280 image resolution. Each image captured is saved in JPG format utilizing not more than 1 MB size. The images are grouped into four folders respective to four species.
- Varghese, Anjli; Jain, Sahil; Prince, A. Amalin; Jawahar, Malathy (2020), Digital Microscopic Image Sensing and Processing for Leather Species Identification, IEEE Sensors Journal, Journal-article, https://doi.org/10.1109/jsen.2020.2991881
- Varghese, Anjli; Jawahar, Malathy; Prince, A. Amalin (2023), Learning species-definite features from digital microscopic leather images, Expert Systems with Applications, Journal-article, https://doi.org/10.1016/j.eswa.2023.119971
- Varghese, Anjli; Jain, Sahil; Jawahar, Malathy; Prince, A. Amalin (2023), Auto-pore segmentation of digital microscopic leather images for species identification, Engineering Applications of Artificial Intelligence, Journal-article, https://doi.org/10.1016/j.engappai.2023.107049