WaterHyacinth: A Comprehensive Image Dataset of Various Water Hyacinth Species From Different Regions of Bangladesh.
Description
The newly created dataset "WaterHyacinth" contains information on numerous kinds of water hyacinth that are often found in different parts of Bangladesh. There are four different species of water hyacinth represented in this dataset. Lemna minor, Eichornia crassipes, Monochoria korsakowii, and Pistia stratiotes are the four separate species. There are 1,790 original images in the entire collection. We gathered these photographs over a period of three months, from July 2023 to September 2023, from separate locations of Sirajganj and Pabna zilla under a variety of outdoor lighting situations as well as inside lighting with acceptable backgrounds. The file also includes 4,050 extra augmented photos. This extensive dataset offers enormous potential for researchers to contribute to environmental science, aquatic science, remote sensing, and water condition monitoring using different Machine Learning and Deep Learning techniques. It is an invaluable asset for future work, including remote sensing with satellite imagery, managing water quality, and conserving biodiversity domains.
Files
Steps to reproduce
Our data was acquired in a number of steps. First of all, we studied about Water Hyacinth and it's different kinds of species that are found in our local areas. When we researched Water Hyacinth, we discovered that it is regarded as the worst aquatic weed in the entire globe. Then, we went on an actual visit to some of the nearby rivers, ponds, and lakes where several varieties of water hyacinth are grown. Following that, we gathered raw photos of these species from several locations. As stated in our data description, we discovered four distinct kinds of water hyacinth in various locations and collected random number of images of them using four different smartphones i.e. Redmi Note 8 Pro, Redmi Note 11, and iPhone 12 and Samsung Galaxy S10. After collecting the pictures, we group them into four folders based on the species. Then, using preprocessing, we remove noisy and ambiguous photos from our dataset. There are a total of 1,790 photos left after preprocessing. We further augmented these images and produced 4,050 extra images to enrich our dataset. These vast dataset will be helpful for researchers to contribute in image classification, various ML, DL techniques.