Bangladeshi Freshwater Fish Image Dataset (BFF-15)
Description
The Bangladeshi Freshwater Fish Image Dataset (BFF-15) is an image dataset that consists of 5577 images of 15 different species of freshwater fish that can be found in Bangladesh. The majority of the images were collected from multiple market locations and a hatchery to ensure good representation of how the fish vary in appearance, how the lighting effects the photos and the types of places that people acquire the photos.
During field visits to local markets and aquaculture operations, the author's personal visits to those locations allowed for manual collection of all photographs. In order for this dataset to support data mining research related to computer vision, deep learning, fish species classification, and aquatic biodiversity monitoring, it is important that these photo collections include samples representative of the species listed above.
Each of the collected images was taken with a smartphone camera or its equivalent (Realme 8 and iPhone 15 Pro) for maximum image detail, meaning that an image resolution will have been preserved through to post-processing steps. Post-processing involved conducting standardisation of photographic images to prepare them for machine learning applications; resizing each of the 5577 images down to square 224 × 224 pixels, removing backgrounds from each image and changing them to a homogeneous white with the intent of reducing noise and improving the accuracy of prediction of species.
The dataset is suitable for tasks such as:
-
Fish species classification
-
Deep learning model training
-
Image recognition research
-
Aquaculture monitoring systems
-
Biodiversity analysis
Dataset Statistics
Total Images: 5,577
Total Species: 15
| Species | Number of Images |
|---|---|
| Grass Carp | 403 |
| Ilish | 233 |
| Katla | 427 |
| Koi | 434 |
| Magur | 282 |
| Mirror Carp | 259 |
| Mrigal | 317 |
| Pabda | 348 |
| Pangas | 510 |
| Puti | 409 |
| Rui | 514 |
| Shol | 296 |
| Silver Carp | 240 |
| Telapia | 383 |
| Tengra | 522 |
Data Collection Locations
Images were collected from the following locations in Bangladesh:
-
Chilahati Bazar (26.246212, 88.794525)
-
BotolGanj Bazar (26.231234, 88.791190)
-
Mirpur-1 Fish Market (23.797606, 90.350655)
-
Ashulia Bazar (23.897087, 90.330834)
-
Bhawlaganj Bazar (26.253909, 88.745152)
-
Chilahati Fish Hatchery (26.245912, 88.801593)
These locations include both local markets and aquaculture sources, ensuring diversity in the collected samples.
Image Acquisition
Images were captured using:
-
Realme 8 smartphone camera
-
iPhone 15 Pro smartphone camera
The photographs were taken directly from freshly harvested fish in market environments and hatchery facilities.
Preprocessing
The following preprocessing steps were applied:
-
Image resizing to 224 × 224 pixels
-
Background removal
-
White background standardization
These preprocessing steps make the dataset suitable for training deep learning models such as CNN, Vision Transformer and Hybrid architectures.
Files
BD-FISH15-CLEAN-DATASET.zip
Files
(33.4 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:ead27d855bb1ec73b8dad5485e5ff509
|
33.4 MB | Preview Download |