Project

Return to Project Catalogue

Collection of fish images to be used in development of autonomous fish identification and sorting tool

Fish passage remains unrealized in Great Lakes tributaries due to the threat of infestation by invasive species and cost of “safe” options such as trap and sort. A selective, autonomously operated passage device using imaged based sorting could provide the cost-effective solution fishery managers need to make fish passage a reality. The first step in developing a fish identification tool however, is collecting images of the fish species to be used by a machine learning approach. Project objectives include: (1.) Obtain ~1000 images of walleye, steelhead, suckers, northern pike, sea lamprey, common carp, Asian carps (silver, black, bighead, as available), and other Great Lakes fishes as available with the Whoosh FishLTM Recognition scanner to begin development of a fish ID classifier. (2.) Create a database containing fish images identified by species, location, and date collected for those species as well as algorithms for identifying sea lamprey images.

Status
Ongoing

GLFC ID
NA

Research Program
FishPass

Research Theme

Start Date
2019

End Date
2020

PI Name
Zielinski, Dan

PI Email
dzielinski@glfc.org

PI Institution
Great Lakes Fishery Commission

Project Keywords



Project Datasets

FishL Low Resolution Images of Fish from the Great Lakes Region

View Dataset

This is a dataset consisting of 6,246 images of fish collected from the Great Lakes region. The images were captured using the FishL(tm) recognition scanner. Before being passed through the scanner, many fish were measured and weighted.

Dataset Format
Scripts and image files

Waterbody
Basin-wide

Variables
image, length, girth, weight

Dataset Keywords

Tools to Apply Deep Convolution Neural Networks to Predict Species in Great Lakes Region

View Dataset

This project contains training and evaluation scripts as well as trained models and generated results for several deep convolutional neural networks (DCNNs). The DCNNs predicted the species of a fish in an image from one of 13 different species of fish from the Great Lakes region. Full details on the evaluation of the classifiers are available in Eickholt et al. 2020. Training and evaluation data are available at FishL Low Resolution Images of Fish from the Great Lakes Region, DOI 10.17605/OSF.IO/KQVG8. Eickholt, J., Kelly, D., Bryan, J., Miehls, S. and Zielinski, D., 2020. Advancements towards selective barrier passage by automatic species identification: applications of deep convolutional neural networks on images of dewatered fish. ICES Journal of Marine Science, 77(7-8), pp.2804-2813. https://doi.org/10.1093/icesjms/fsaa150.

Dataset Format
Python scripts

Waterbody
Basin-wide

Variables
image, length, girth, weight

Dataset Keywords

Image and biometric data for fish from Great Lakes tributaries collected during spring 2019

View Dataset

Image and biometric data were collected for 22 species of fish from Great Lakes Tributaries in Michigan and Ohio, and the Illinois River for the purpose of developing a fish identification classifier. Data consists of a comma delimited spreadsheet that identifies image file names and associated fish identification number, common name, species code, family name, genus, and species, date collected, river from which each fish was collected, location of sampling, fish fork length in millimeters, girth in millimeters, weight in kilograms, and personnel involved with image collection. Biometric data are saved as .csv comma delimited format and image files are saved as .png file type.

Dataset Format
.csv .png

Waterbody
Basin-wide

Variables
image, length, girth, weight

Dataset Keywords



Other Project Products