DriedFishBD: A Detailed Image Dataset of Common Small-sized Dried Fish Varieties in Bangladesh for Identification and Classification.
Description
Type of data: 1024 x 1024 px images of small-sized dried fishes. Data format: JPEG. Contents of the dataset: Images of different varieties of small-sized dried fishes in Bangladesh from (i) single fish, (ii) bulk fish, (iii) fish head, and (iv) fish tail perspectives. Number of classes: Ten small-sized fish varieties – (1) Chanda, (2) Chapila, (3) Chela, (4) Chepa, (5) Guchi, (6) Kachki, (7) Loitta, and (8) Tengra. Number of images: (A) Total images in the Original Dataset = 3,288, and (B) Total images in the Augmented Dataset = 39,456. Distribution of instances: (A) Number of images in the Original Dataset: (1) Chanda: 400 images in total. Single = 100, Bulk = 100, Head = 100, Tail = 100. (2) Chapila: 422 images in total. Single = 120, Bulk = 100, Head = 102, Tail = 100. (3) Chela: 400 images in total. Single = 100, Bulk = 100, Head = 100, Tail = 100. (4) Chepa: 400 images in total. Single = 100, Bulk = 100, Head = 100, Tail = 100. (5) Guchi: 427 images in total. Single = 105, Bulk = 108, Head = 103, Tail = 111. (6) Kachki: 435 images in total. Single = 121, Bulk = 107, Head = 103, Tail = 104. (7) Loitta: 400 images in total. Single = 100, Bulk = 100, Head = 100, Tail = 100. (8) Tengra: 404 images in total. Single = 100, Bulk = 100, Head = 100, Tail = 104. (B) Number of images in the Augmented Dataset: (1) Chanda: 4,800 images in total. Single = 1,200, Bulk = 1,200, Head = 1,200, Tail = 1,200. (2) Chapila: 5,064 images in total. Single = 1,440, Bulk = 1,200, Head = 1,224, Tail = 1,200. (3) Chela: 4,800 images in total. Single = 1,200, Bulk = 1,200, Head = 1,200, Tail = 1,200. (4) Chepa: 4,800 images in total. Single = 1,200, Bulk = 1,200, Head = 1,200, Tail = 1,200. (5) Guchi: 5,124 images in total. Single = 1,260, Bulk = 1,296, Head = 1,236, Tail = 1,332. (6) Kachki: 5,220 images in total. Single = 1,452, Bulk = 1,284, Head = 1,236, Tail = 1,248. (7) Loitta: 4,800 images in total. Single = 1,200, Bulk = 1,200, Head = 1,200, Tail = 1,200. (8) Tengra: 4,848 images in total. Single = 1,200, Bulk = 1,200, Head = 1,200, Tail = 1,248. Dataset size: (A) Total size of Original Dataset is 403 MB and compressed ZIP file size is 390 MB. (B) Total size of Augmented Dataset is 3.23 GB and compressed ZIP size is 3.01 GB. Data acquisition process: Images of various small-sized dried fish categories are captured through a high-definition smartphone camera focusing from different angles. Data source location: Local dried fish markets located in Dhaka, Bangladesh. Where applicable: Training and evaluating machine learning and deep learning models to identify and classify small-sized dried fish varieties in Bangladesh which can be useful in Food Processing Industry, Fisheries Management, Supply Chain and Logistics, Retail and E-Commerce, Food Safety and Regulatory Compliance, Environmental and Conservation Research, Ethnography and Cultural Studies, Nutritional Science, Artificial Intelligence and Machine Learning, and Agricultural Economics.