Abstract

Title: Standardizing the Process of Waterfowl Species Identification using Neural Networks

Authors: Andres Segura and Jose Vega (undergraduate students in CS)

Jonatan Contreras, Sara Gonzalez (PhD students in CS and Biology respectively)

Philip Lavretsky, Martine Ceberio (faculty advisors)

Abstract:

Waterfowl are socioeconomically important worldwide and are subject of constant scientific analysis. From banding operations to the key aspects of adaptive harvest management models, their phenotype is used to evaluate their sex / age, to monitor population. Properly evaluating phenotype is critical for many aspects of waterfowl conservation. The mallard is a Holarctic species and has been used as food since humanity’s early hunting-gathering times. Through time, it has diverged into other species by natural movement and isolation from other populations. However, evolution is a long process, so these new species can be considered infants in the scope of evolutionary time. This leads to ancestral trait retention where new species still show much of the same traits as their ancestor. Caused by a lack of sturdy reproductive barriers, this can lead to confounding visual traits used when identifying species. In a recent study by Lavretsky et. al. (2019), “20% of phenotypically identified mallards and black ducks were incorrect…” It is crucial to standardize against identification error to form more informed ecological decisions.

This study aims to automate phenotypic recognition by using ML and computer vision. Our work is a first step in creating a system that will identify the chosen species on the field with a high degree of accuracy to support the identification of waterfowl. The system is trained with a data set from the UTEP Lavretsky Lab, complete with genetically verified species to ensure high quality, pre-labeled images. Preliminary results are discussed, and future work presented.

Links

Mallard DNA found in 5,700 year old ‘chewing gum’:
https://www.nationalgeographic.com/history/article/dna-stone-age-chewing-gum-microbiome-story

Lavretsky, Philip et al. “Identifying hybrids & the genomics of hybridization: Mallards & American black ducks of Eastern North America.” Ecology and evolution vol. 9,6 3470-3490. 27 Feb. 2019, doi:10.1002/ece3.4981
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6434578/

Sara Gonzales’s presentation for The 27th Joint NMSU/UTEP Workshop on Mathematics, Computer Science, and Computational Sciences
https://docs.google.com/presentation/d/1_sH-0FHWP6lTK4JLVZKIW2aDrC4HljR_xT-pdilmgO0/edit?usp=sharing

A comprehensive guide to convolutional neural networks
https://towardsdatascience.com/a-comprehensive-guide-to-convolutional-neural-networks-the-eli5-way-3bd2b1164a53

• CONTACT INFORMATION
• Sara Gonzalez (sgonzalez28 [at] miners.utep.edu)
• Jonatan Contreras (jmcontreras2 [at] miners.utep.edu)