Title: | A deep learning method for visual recognition of snake species |
Authors: | Chamidullin, Rail Šulc, Milan Matas, Jiří Picek, Lukáš |
Citation: | CHAMIDULLIN, R. ŠULC, M. MATAS, J. PICEK, L. A deep learning method for visual recognition of snake species. In Proceedings of the Working Notes of CLEF 2021 - Conference and Labs of the Evaluation Forum. neuvedeno: CEUR-WS, 2021. s. 1512-1525. ISBN: neuvedeno , ISSN: 1613-0073 |
Issue Date: | 2021 |
Publisher: | CEUR-WS |
Document type: | konferenční příspěvek ConferenceObject |
URI: | 2-s2.0-85113481716 http://hdl.handle.net/11025/47277 |
ISBN: | neuvedeno |
ISSN: | 1613-0073 |
Keywords in different language: | Snake species identification;Fine-grained classification;Computer vision;Convolutional neural networks;Deep learning |
Abstract in different language: | The paper presents a method for image-based snake species identification. The proposed method is based on deep residual neural networks - ResNeSt, ResNeXt and ResNet - fine-tuned from ImageNet pre-trained checkpoints. We achieve performance improvements by: discarding predictions of species that do not occur in the country of the query; combining predictions from an ensemble of classifiers; and applying mixed precision training, which allows training neural networks with larger batch size. We experimented with loss functions inspired by the considered metrics: soft F1 loss and weighted cross entropy loss. However, the standard cross entropy loss achieved superior results both in accuracy and in F1 measures. The proposed method scored third in the SnakeCLEF 2021 challenge, achieving 91.6% classification accuracy, Country F1 Score of 0.860, and F1 Score of 0.830. |
Rights: | © authors |
Appears in Collections: | Konferenční příspěvky / Conference Papers (KKY) OBD |
Files in This Item:
File | Size | Format | |
---|---|---|---|
Chamidullin_A_deep_learning_paper-128_2021.pdf | 918,87 kB | Adobe PDF | View/Open |
Please use this identifier to cite or link to this item:
http://hdl.handle.net/11025/47277
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.