BDFreshFish: A Comprehensive Image Dataset for Machine Learning Applications on Bangladeshi Freshwater Fishes

Published: 10 January 2024| Version 1 | DOI: 10.17632/29kjy99kkh.1
Contributors:
,
,
,

Description

The "BDFreshFish" dataset contains a set of image data for eight different categories of native freshwater fish from various parts of Bangladesh. The eight distinct classes of fish have the following scientific names: Anabas Testudineus, Batasio Tengana, Channa Punctata, Marcrobrachium malcoimsonii, Heteropneustes Fossilis, Mastacembelus Armatus, Ompok Bimaculatus, and Puntius Sophore. Our collection comprises 3100+ images across these eight classes, with 300 augmented images and approximately 100 raw images in each class. The majority of raw images originate from rivers in Bangladesh. The dataset has four primary fish species characteristics: head, body, scales, and fins. Each original image was taken in adequate natural light against the right backdrop. Each image was properly organized into an appropriate folder, allowing different machine learning and deep learning models to make the best use of the images. Utilizing this enormous dataset and various traditional machine learning (TML) and deep learning (DL) techniques, researchers could make significant advancements in analyzing particular fish category classifications to analyze the freshwater environment and fish movement.

Files

Steps to reproduce

The data collection includes a set of procedures: 1. Fish Species Selection. 2. Image Capturing. 3. Data Curing / Data cleaning. 4. Image Augmentation. 5. Image Compression and Storing. 6. Fish Classification with Deep Learning.

Institutions

Khwaja Yunus Ali University

Categories

Machine Learning, Fisheries Co-Management, Freshwater Fisheries

Licence