1
|
Garga B, Abboubakar H, Sourpele RS, Gwet DLL, Bitjoka L. Pollen Grain Classification Using Some Convolutional Neural Network Architectures. J Imaging 2024; 10:158. [PMID: 39057729 PMCID: PMC11277931 DOI: 10.3390/jimaging10070158] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 06/05/2024] [Revised: 06/24/2024] [Accepted: 06/26/2024] [Indexed: 07/28/2024] Open
Abstract
The main objective of this work is to use convolutional neural networks (CNN) to improve the performance in previous works on their baseline for pollen grain classification, by improving the performance of the following eight popular architectures: InceptionV3, VGG16, VGG19, ResNet50, NASNet, Xception, DenseNet201 and InceptionResNetV2, which are benchmarks on several classification tasks, like on the ImageNet dataset. We use a well-known annotated public image dataset for the Brazilian savanna, called POLLEN73S, composed of 2523 images. Holdout cross-validation is the name of the method used in this work. The experiments carried out showed that DenseNet201 and ResNet50 outperform the other CNNs tested, achieving results of 97.217% and 94.257%, respectively, in terms of accuracy, higher than the existing results, with a difference of 1.517% and 0.257%, respectively. VGG19 is the architecture with the lowest performance, achieving a result of 89.463%.
Collapse
Affiliation(s)
- Benjamin Garga
- ENSAI, Laboratory of Energy, Signal, Imaging and Automation, University of Ngaoundere, Ngaoundere P.O. Box 455, Cameroon; (B.G.); (R.S.S.); (D.L.L.G.); (L.B.)
| | - Hamadjam Abboubakar
- ENSAI, Laboratory of Energy, Signal, Imaging and Automation, University of Ngaoundere, Ngaoundere P.O. Box 455, Cameroon; (B.G.); (R.S.S.); (D.L.L.G.); (L.B.)
- Departement of Computer Engineering, University Institute of Technology, University of Ngaoundere, Ngaoundere P.O. Box 455, Cameroon
- Laboratory of Analysis, Simulations and Tests, University of Ngaoundere, Ngaoundere P.O. Box 455, Cameroon
| | - Rodrigue Saoungoumi Sourpele
- ENSAI, Laboratory of Energy, Signal, Imaging and Automation, University of Ngaoundere, Ngaoundere P.O. Box 455, Cameroon; (B.G.); (R.S.S.); (D.L.L.G.); (L.B.)
- ENSAI, Department of Mathematics and Computer Sciences, University of Ngaoundere, Ngaoundere P.O. Box 455, Cameroon
| | - David Libouga Li Gwet
- ENSAI, Laboratory of Energy, Signal, Imaging and Automation, University of Ngaoundere, Ngaoundere P.O. Box 455, Cameroon; (B.G.); (R.S.S.); (D.L.L.G.); (L.B.)
- EGCIM, Department of Fundamental Sciences and Engineering Techniques, University of Ngaoundere, Ngaoundere P.O. Box 455, Cameroon
| | - Laurent Bitjoka
- ENSAI, Laboratory of Energy, Signal, Imaging and Automation, University of Ngaoundere, Ngaoundere P.O. Box 455, Cameroon; (B.G.); (R.S.S.); (D.L.L.G.); (L.B.)
- ENSAI, Department of Electrical Engineering, Energetics and Automation, University of Ngaoundere, Ngaoundere P.O. Box 455, Cameroon
| |
Collapse
|
2
|
Li C, Polling M, Cao L, Gravendeel B, Verbeek FJ. Analysis of Automatic Image Classification Methods for Urticaceae Pollen Classification. Neurocomputing 2022. [DOI: 10.1016/j.neucom.2022.11.042] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/27/2022]
|
3
|
Automatic Classification of Pollen Grain Microscope Images Using a Multi-Scale Classifier with SRGAN Deblurring. APPLIED SCIENCES-BASEL 2022. [DOI: 10.3390/app12147126] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Track Full Text] [Subscribe] [Scholar Register] [Indexed: 02/05/2023]
Abstract
Pollen allergies are seasonal epidemic diseases that are accompanied by high incidence rates, especially in Beijing, China. With the development of deep learning, key progress has been made in the task of automatic pollen grain classification, which could replace the time-consuming and laborious manual identification process using a microscope. In China, few pioneering works have made significant progress in automatic pollen grain classification. Therefore, we first constructed a multi-class and large-scale pollen grain dataset for the Beijing area in preparation for the task of pollen classification. Then, a deblurring pipeline was designed to enhance the quality of the pollen grain images selectively. Moreover, as pollen grains vary greatly in size and shape, we proposed an easy-to-implement and efficient multi-scale deep learning architecture. Our experimental results showed that our architecture achieved a 97.7% accuracy, based on the Resnet-50 backbone network, which proved that the proposed method could be applied successfully to the automatic identification of pollen grains in Beijing.
Collapse
|
4
|
Khanzhina N, Filchenkov A, Minaeva N, Novoselova L, Petukhov M, Kharisova I, Pinaeva J, Zamorin G, Putin E, Zamyatina E, Shalyto A. Combating data incompetence in pollen images detection and classification for pollinosis prevention. Comput Biol Med 2022; 140:105064. [PMID: 34861642 DOI: 10.1016/j.compbiomed.2021.105064] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 10/01/2021] [Revised: 11/20/2021] [Accepted: 11/20/2021] [Indexed: 11/30/2022]
Abstract
Automatic pollen images recognition is crucial for pollinosis symptoms prevention and treatment. The problem of pollen recognition can be efficiently solved using deep learning, however neural networks require tens of thousands of images to generalize. At the same time, the existing open pollen images datasets are very small. In this paper, we present a novel open pollen dataset annotated for both detection and classification tasks. Based on our dataset we study learning from a small data using different state-of-the-art approaches. For the detection task we propose to use our new Bayesian RetinaNet network, which models aleatoric uncertainty. We compare it with the baseline RetinaNet and demonstrate that our model allows for higher detection precision. For the classification task we compare the impact of pre-training on the synthetic images from generative adversarial networks (GANs) and metric-based few-shot learning. Namely, we pre-trained our small convolutional neural network and Siamese neural network classifiers on synthetic pollen images generated by two GANs: StyleGAN and Self-attention GAN. The best classifier is the convolutional neural network pre-trained on StyleGAN images. Our best models achieved 96.3% of mean average precision for the detection task and 97.7% of F1 measure for the classification task on 13 pollen plant species. We have implemented the best models in our pollen recognition web service, which is available for palynologists on request.
Collapse
Affiliation(s)
- Natalia Khanzhina
- Machine Learning Lab, ITMO University, 49 Kronverksky Pr., Saint Petersburg, 197 101, Russia.
| | - Andrey Filchenkov
- Machine Learning Lab, ITMO University, 49 Kronverksky Pr., Saint Petersburg, 197 101, Russia
| | - Natalia Minaeva
- Perm State Medical University, 26 Petropavlovskaya St., Perm, 614 000, Russia
| | - Larisa Novoselova
- Perm State National Research University, 15 Bukireva St., Perm, 614 990, Russia
| | - Maxim Petukhov
- Machine Learning Lab, ITMO University, 49 Kronverksky Pr., Saint Petersburg, 197 101, Russia
| | - Irina Kharisova
- Perm State National Research University, 15 Bukireva St., Perm, 614 990, Russia
| | - Julia Pinaeva
- Perm State National Research University, 15 Bukireva St., Perm, 614 990, Russia
| | - Georgiy Zamorin
- Perm State National Research University, 15 Bukireva St., Perm, 614 990, Russia
| | - Evgeny Putin
- Insilico Medicine Hong Kong Ltd, Pak Shek Kok, New Territories, Hong Kong
| | - Elena Zamyatina
- Perm State National Research University, 15 Bukireva St., Perm, 614 990, Russia; National Research University "Higher School of Economics", Faculty of Economics, Management, and Business Informatics, 38 Studencheskaya St., 614 070, Perm, Russia
| | - Anatoly Shalyto
- Machine Learning Lab, ITMO University, 49 Kronverksky Pr., Saint Petersburg, 197 101, Russia
| |
Collapse
|
5
|
Pospiech M, Javůrková Z, Hrabec P, Štarha P, Ljasovská S, Bednář J, Tremlová B. Identification of pollen taxa by different microscopy techniques. PLoS One 2021; 16:e0256808. [PMID: 34469471 PMCID: PMC8409677 DOI: 10.1371/journal.pone.0256808] [Citation(s) in RCA: 3] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Abstract] [MESH Headings] [Grants] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 04/18/2021] [Accepted: 08/17/2021] [Indexed: 11/28/2022] Open
Abstract
Melissopalynology is an important analytical method to identify botanical origin of honey. Pollen grain recognition is commonly performed by visual inspection by a trained person. An alternative method for visual inspection is automated pollen analysis based on the image analysis technique. Image analysis transfers visual information to mathematical descriptions. In this work, the suitability of three microscopic techniques for automatic analysis of pollen grains was studied. 2D and 3D morphological characteristics, textural and colour features, and extended depth of focus characteristics were used for the pollen discrimination. In this study, 7 botanical taxa and a total of 2482 pollen grains were evaluated. The highest correct classification rate of 93.05% was achieved using the phase contrast microscopy, followed by the dark field microscopy reaching 91.02%, and finally by the light field microscopy reaching 88.88%. The most significant discriminant characteristics were morphological (2D and 3D) and colour characteristics. Our results confirm the potential of using automatic pollen analysis to discriminate pollen taxa in honey. This work provides the basis for further research where the taxa dataset will be increased, and new descriptors will be studied.
Collapse
Affiliation(s)
- Matej Pospiech
- Faculty of Veterinary Hygiene and Ecology, Department of Plant Origin Food Sciences, University of Veterinary Sciences Brno, Brno, Czech Republic
| | - Zdeňka Javůrková
- Faculty of Veterinary Hygiene and Ecology, Department of Plant Origin Food Sciences, University of Veterinary Sciences Brno, Brno, Czech Republic
- * E-mail:
| | - Pavel Hrabec
- Faculty of Mechanical Engineering, Department of Statistics and Optimization, Brno University of Technology, Brno, Czech Republic
| | - Pavel Štarha
- Faculty of Mechanical Engineering, Department of Computer Graphics and Geometry, Brno University of Technology, Brno, Czech Republic
| | - Simona Ljasovská
- Faculty of Veterinary Hygiene and Ecology, Department of Plant Origin Food Sciences, University of Veterinary Sciences Brno, Brno, Czech Republic
| | - Josef Bednář
- Faculty of Mechanical Engineering, Department of Statistics and Optimization, Brno University of Technology, Brno, Czech Republic
| | - Bohuslava Tremlová
- Faculty of Veterinary Hygiene and Ecology, Department of Plant Origin Food Sciences, University of Veterinary Sciences Brno, Brno, Czech Republic
| |
Collapse
|
6
|
Kubik-Komar A, Kubera E, Piotrowska-Weryszko K. Selection of morphological features of pollen grains for chosen tree taxa. Biol Open 2018; 7:bio.031237. [PMID: 29643087 PMCID: PMC5992530 DOI: 10.1242/bio.031237] [Citation(s) in RCA: 4] [Impact Index Per Article: 0.7] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/20/2022] Open
Abstract
The basis of aerobiological studies is to monitor airborne pollen concentrations and pollen season timing. This task is performed by appropriately trained staff and is difficult and time consuming. The goal of this research is to select morphological characteristics of grains that are the most discriminative for distinguishing between birch, hazel and alder taxa and are easy to determine automatically from microscope images. This selection is based on the split attributes of the J4.8 classification trees built for different subsets of features. Determining the discriminative features by this method, we provide specific rules for distinguishing between individual taxa, at the same time obtaining a high percentage of correct classification. The most discriminative among the 13 morphological characteristics studied are the following: number of pores, maximum axis, minimum axis, axes difference, maximum oncus width, and number of lateral pores. The classification result of the tree based on this subset is better than the one built on the whole feature set and it is almost 94%. Therefore, selection of attributes before tree building is recommended. The classification results for the features easiest to obtain from the image, i.e. maximum axis, minimum axis, axes difference, and number of lateral pores, are only 2.09 pp lower than those obtained for the complete set, but 3.23 pp lower than the results obtained for the selected most discriminating attributes only.
Collapse
Affiliation(s)
- Agnieszka Kubik-Komar
- University of Life Sciences in Lublin, Department of Applied Mathematics and Computer Science, Akademicka 13, 20-950 Lublin, Poland
| | - Elżbieta Kubera
- University of Life Sciences in Lublin, Department of Applied Mathematics and Computer Science, Akademicka 13, 20-950 Lublin, Poland
| | | |
Collapse
|
7
|
|