Evaluating the performance of artificial neural networks for the classification of freshwater benthic macroinvertebrates

Joutsijoki, Henry; Meissner, Kristian; Gabbouj, Moncef; Kıranyaz, Serkan; Raitoharju, Jenni; Ärje, Johanna; Kärkkäinen, Salme; Tirronen, Ville; Turpeinen, Tuomas; Juhola, Martti

doi:10.1016/j.ecoinf.2014.01.004

Cited by 31 publications

(33 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Image-based species identification methods on arthropods have been applied with success on samples in the laboratory (Joutsijoki et al, 2014). In order to fully implement the advantages of camera traps, there is a need for implementing image classification techniques to automatically identify and recognize species (Weinstein, 2017).…”

Section: Introductionmentioning

confidence: 99%

Species‐level image classification with convolutional neural network enables insect identification from habitus images

Hansen

Svenning

Olsen

et al. 2019

Ecology and Evolution

View full text Add to dashboard Cite

1. Changes in insect biomass, abundance, and diversity are challenging to track at sufficient spatial, temporal, and taxonomic resolution. Camera traps can capture habitus images of ground-dwelling insects. However, currently sampling involves manually detecting and identifying specimens. Here, we test whether a convolutional neural network (CNN) can classify habitus images of ground beetles to species level, and estimate how correct classification relates to body size, number of species inside genera, and species identity.2. We created an image database of 65,841 museum specimens comprising 361 carabid beetle species from the British Isles and fine-tuned the parameters of a pretrained CNN from a training dataset. By summing up class confidence values within genus, tribe, and subfamily and setting a confidence threshold, we trade-off between classification accuracy, precision, and recall and taxonomic resolution.3. The CNN classified 51.9% of 19,164 test images correctly to species level and 74.9% to genus level. Average classification recall on species level was 50.7%.Applying a threshold of 0.5 increased the average classification recall to 74.6% at the expense of taxonomic resolution. Higher top value from the output layer and larger sized species were more often classified correctly, as were images of species in genera with few species. 4. Fine-tuning enabled us to classify images with a high mean recall for the whole test dataset to species or higher taxonomic levels, however, with high variability.This indicates that some species are more difficult to identify because of properties such as their body size or the number of related species.5. Together, species-level image classification of arthropods from museum collections and ecological monitoring can substantially increase the amount of occurrence data that can feasibly be collected. These tools thus provide new opportunities in understanding and predicting ecological responses to environmental change. Jens-Christian Svenninghttps://orcid.

show abstract

Section: Introductionmentioning

confidence: 99%

Species‐level image classification with convolutional neural network enables insect identification from habitus images

Hansen

Svenning

Olsen

et al. 2019

Ecology and Evolution

View full text Add to dashboard Cite

show abstract

“…citizen-science monitoring (Dickinson et al, 2012) and automated identification methods (e.g. Blaschko et al, 2005;Culverhouse et al, 2006;Lytle et al, 2010;Kiranyaz et al, 2011;Ärje et al, 2013;Joutsijoki et al, 2014). However, such approaches may introduce additional bias and variation into biological indices calculated from samples due to identification errors.…”

Section: Introductionmentioning

confidence: 99%

The effect of automated taxa identification errors on biological indices

rje

Krkkinen

Meissner

et al. 2017

Expert Systems with Applications

Self Cite

View full text Add to dashboard Cite

show abstract

“…Hence, the automation of the identification process would cut costs greatly. Often the identification of benthic macroinvertebrates can be routine work for human experts (Joutsijoki et al, 2014). Thus, the automation of this process could relieve the workload of taxonomists to solve some other more specialized problems (Joutsijoki et al, 2014).…”

Section: Introductionmentioning

confidence: 99%

“…Often the identification of benthic macroinvertebrates can be routine work for human experts (Joutsijoki et al, 2014). Thus, the automation of this process could relieve the workload of taxonomists to solve some other more specialized problems (Joutsijoki et al, 2014). An automated identification process would also enable biologists to collect a larger numbers of samples, which is recommended when benthic macroinvertebrates are used in biomonitoring.…”

Section: Introductionmentioning

confidence: 99%

“…An automated identification process would also enable biologists to collect a larger numbers of samples, which is recommended when benthic macroinvertebrates are used in biomonitoring. However, the need of human-made identification cannot be totally removed since human expertise is required when constructing a training set for the use of classification methods (Joutsijoki et al, 2014).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A comparison of classification methods in automated taxa identification of benthic macroinvertebrates

Joutsijoki

Juhola

2017

IJDS

Self Cite

View full text Add to dashboard Cite

In this research we examined the automated taxa identification of benthic macroinvertebrates. Benthic macroinvertebrates play an important role in biomonitoring. They can be used in water quality assessments. Identification of benthic macroinvertebrates is made usually by highly trained experts, but this approach has high costs and, hence, the automation of this identification process could reduce the costs and would make wider biomonitoring possible. The automated taxa identification of benthic macroinvertebrates returns to image classification. We applied altogether 11 different classification methods to the image dataset of eight taxonomic groups of benthic macroinvertebrates. Wide experimental tests were performed. The best results, around 94% accuracies, were achieved when Quadratic Discriminant Analysis, Radial Basis Function network and Multi-Layer Perceptron were used. On the basis of the results, it can be said that the automated taxa identification of benthic macroinvertebrates is possible with high accuracy.

show abstract

Evaluating the performance of artificial neural networks for the classification of freshwater benthic macroinvertebrates

Cited by 31 publications

References 30 publications

Species‐level image classification with convolutional neural network enables insect identification from habitus images

Species‐level image classification with convolutional neural network enables insect identification from habitus images

The effect of automated taxa identification errors on biological indices

A comparison of classification methods in automated taxa identification of benthic macroinvertebrates

Contact Info

Product

Resources

About