Applications of deep convolutional neural networks to digitized natural history collections
Natural history collections contain data that are critical for many scientific endeavors. Recent efforts in mass digitization are generating large datasets from these collections that can provide unprecedented insight. Here, we present examples of how deep convolutional neural networks can be applie...
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Pensoft Publishers
2017-11-01
|
| Series: | Biodiversity Data Journal |
| Subjects: | |
| Online Access: | https://bdj.pensoft.net/articles.php?id=21139 |
| _version_ | 1819024504023678976 |
|---|---|
| author | Eric Schuettpelz Paul Frandsen Rebecca Dikow Abel Brown Sylvia Orli Melinda Peters Adam Metallo Vicki Funk Laurence Dorr |
| author_facet | Eric Schuettpelz Paul Frandsen Rebecca Dikow Abel Brown Sylvia Orli Melinda Peters Adam Metallo Vicki Funk Laurence Dorr |
| author_sort | Eric Schuettpelz |
| collection | DOAJ |
| description | Natural history collections contain data that are critical for many scientific endeavors. Recent efforts in mass digitization are generating large datasets from these collections that can provide unprecedented insight. Here, we present examples of how deep convolutional neural networks can be applied in analyses of imaged herbarium specimens. We first demonstrate that a convolutional neural network can detect mercury-stained specimens across a collection with 90% accuracy. We then show that such a network can correctly distinguish two morphologically similar plant families 96% of the time. Discarding the most challenging specimen images increases accuracy to 94% and 99%, respectively. These results highlight the importance of mass digitization and deep learning approaches and reveal how they can together deliver powerful new investigative tools. |
| first_indexed | 2024-12-21T04:55:50Z |
| format | Article |
| id | doaj.art-c48fc52b11444671ae24670f76cabb39 |
| institution | Directory Open Access Journal |
| issn | 1314-2836 1314-2828 |
| language | English |
| last_indexed | 2024-12-21T04:55:50Z |
| publishDate | 2017-11-01 |
| publisher | Pensoft Publishers |
| record_format | Article |
| series | Biodiversity Data Journal |
| spelling | doaj.art-c48fc52b11444671ae24670f76cabb392022-12-21T19:15:22ZengPensoft PublishersBiodiversity Data Journal1314-28361314-28282017-11-0151910.3897/BDJ.5.e2113921139Applications of deep convolutional neural networks to digitized natural history collectionsEric Schuettpelz0Paul Frandsen1Rebecca Dikow2Abel Brown3Sylvia Orli4Melinda Peters5Adam Metallo6Vicki Funk7Laurence Dorr8National Museum of Natural History, Smithsonian InstitutionOffice of the Chief Information Officer, Smithsonian InstitutionOffice of the Chief Information Officer, Smithsonian InstitutionNVIDIANational Museum of Natural History, Smithsonian InstitutionNational Museum of Natural History, Smithsonian InstitutionOffice of the Chief Information Officer, Smithsonian InstitutionNational Museum of Natural History, Smithsonian InstitutionNational Museum of Natural History, Smithsonian InstitutionNatural history collections contain data that are critical for many scientific endeavors. Recent efforts in mass digitization are generating large datasets from these collections that can provide unprecedented insight. Here, we present examples of how deep convolutional neural networks can be applied in analyses of imaged herbarium specimens. We first demonstrate that a convolutional neural network can detect mercury-stained specimens across a collection with 90% accuracy. We then show that such a network can correctly distinguish two morphologically similar plant families 96% of the time. Discarding the most challenging specimen images increases accuracy to 94% and 99%, respectively. These results highlight the importance of mass digitization and deep learning approaches and reveal how they can together deliver powerful new investigative tools.https://bdj.pensoft.net/articles.php?id=21139convolutional neural networksdeep learningm |
| spellingShingle | Eric Schuettpelz Paul Frandsen Rebecca Dikow Abel Brown Sylvia Orli Melinda Peters Adam Metallo Vicki Funk Laurence Dorr Applications of deep convolutional neural networks to digitized natural history collections Biodiversity Data Journal convolutional neural networks deep learning m |
| title | Applications of deep convolutional neural networks to digitized natural history collections |
| title_full | Applications of deep convolutional neural networks to digitized natural history collections |
| title_fullStr | Applications of deep convolutional neural networks to digitized natural history collections |
| title_full_unstemmed | Applications of deep convolutional neural networks to digitized natural history collections |
| title_short | Applications of deep convolutional neural networks to digitized natural history collections |
| title_sort | applications of deep convolutional neural networks to digitized natural history collections |
| topic | convolutional neural networks deep learning m |
| url | https://bdj.pensoft.net/articles.php?id=21139 |
| work_keys_str_mv | AT ericschuettpelz applicationsofdeepconvolutionalneuralnetworkstodigitizednaturalhistorycollections AT paulfrandsen applicationsofdeepconvolutionalneuralnetworkstodigitizednaturalhistorycollections AT rebeccadikow applicationsofdeepconvolutionalneuralnetworkstodigitizednaturalhistorycollections AT abelbrown applicationsofdeepconvolutionalneuralnetworkstodigitizednaturalhistorycollections AT sylviaorli applicationsofdeepconvolutionalneuralnetworkstodigitizednaturalhistorycollections AT melindapeters applicationsofdeepconvolutionalneuralnetworkstodigitizednaturalhistorycollections AT adammetallo applicationsofdeepconvolutionalneuralnetworkstodigitizednaturalhistorycollections AT vickifunk applicationsofdeepconvolutionalneuralnetworkstodigitizednaturalhistorycollections AT laurencedorr applicationsofdeepconvolutionalneuralnetworkstodigitizednaturalhistorycollections |