Amur Tiger Individual Identification Based on the Improved InceptionResNetV2
<p>Rectangle box loss, objectness loss, and classification loss in the training proces.</p> "> Figure 2
<p>Validation indexes. (<b>a</b>) confusion matrix diagram; (<b>b</b>) the P-R curve. The P-R curve colors represent the following categories: Light Blue line - head, Orange line - left body, Green line - right body, Deep Blue line - all classes.</p> "> Figure 3
<p>The actual annotations of the validation dataset.</p> "> Figure 4
<p>The annotation results of the validation dataset by the object detection model.</p> "> Figure 5
<p>An example of segmentation results.</p> "> Figure 6
<p>The structure of the Inception-ResNetV2 module.</p> "> Figure 7
<p>The structure of the improved InceptionResNetV2 model.</p> "> Figure 8
<p>Image padding based on the long side.</p> "> Figure 9
<p>Data enhancement. (<b>a</b>) reduce brightness; (<b>b</b>) increase brightness; (<b>c</b>) reduce contrast; (<b>d</b>) increase contrast; (<b>e</b>) hue adjustment; (<b>f</b>) saturation adjustment; (<b>g</b>) blur image.</p> "> Figure 10
<p>Model training process.</p> "> Figure 11
<p>Comparison of training effectiveness with different dropout rates. (<b>a</b>) dropout rate = 0; (<b>b</b>) dropout rate = 0.1; (<b>c</b>) dropout rate = 0.2; (<b>d</b>) dropout rate = 0.3; (<b>e</b>) dropout rate = 0.4; (<b>f</b>) dropout rate = 0.5.</p> "> Figure 12
<p>Comparison of training effectiveness with different attention mechanisms. (<b>a</b>) Adding an SE module; (<b>b</b>) adding an ECA module; (<b>c</b>) Adding an CBAM module.</p> "> Figure 13
<p>Comparison of the accuracy of different models on the same test set.</p> "> Figure 14
<p>Comparison of the loss of different models on the same test set.</p> ">
Abstract
:Simple Summary
Abstract
1. Introduction
2. Amur Tiger Stripe Part Detection with YOLOV5
2.1. Material and Data Preparation
2.2. Stripe Part Detection and Result Analysis
- (1)
- Number of iterations: epoch = 300;
- (2)
- Batch size: batch_size = 16;
- (3)
- Initial learning rate: lr0 = 0.01 (dynamically reduce the learning rate using a cosine function)
- (4)
- Optimizer: optimizer = SGD
2.2.1. Training Loss
2.2.2. Validation Index
2.2.3. Detection and Segmentation Results
2.2.4. Performance Comparison with Other Object Detection Models
3. The Improved InceptionResNetV2 Model
3.1. Selection of the Base Network
3.2. Improved InceptionResNetV2 Model
3.2.1. Methodology
- Adding Dropout Layer
- 2.
- Introduction of CBAM
3.2.2. Definition of the Improved InceptionResNetV2 Model
4. Experimental Results and Analysis
4.1. Data Preprocessing
4.2. Pre-Trained Model Loading
4.3. Model Training
4.3.1. Experimental Environment Setup and Training Parameters
- (1)
- Number of iterations: epoch = 20;
- (2)
- Batch size: batch_size = 16;
- (3)
- Initial learning rate: lr0 = 0.001;
- (4)
- optimizer: optimizer = Adam;
- (5)
- loss function: loss = Cross-entropy.
4.3.2. Model Training Process
4.4. Comparison of Results
4.4.1. Comparison and Analysis of Different Dropout Rates
4.4.2. Comparison and Analysis of Training Effectiveness with Different Attention Mechanisms
4.4.3. Comparison with Other Models
4.4.4. Comparison and Analysis of Different Physiological Characteristic Parts
4.5. Discussion
5. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Acknowledgments
Conflicts of Interest
References
- Jiang, Y.; Tian, J.; Zhao, J.; Tang, X. The connotation and assessment framework of national park ecosystem integrity: A case study of the Amur Tiger and Leopard National Park. Biodivers. Sci. 2021, 29, 1279. [Google Scholar] [CrossRef]
- Zheng, Z.; Huang, L. Application Status and Prospect of Computer Vision Technology in Rare Wild Animal Disease Monitoring and Early Warning. Prog. Vet. Med. 2024, 45, 118–125. [Google Scholar]
- He, Y. Research on Detection and Individual Recognition of Giant Pandas Based on Convolutional Neural Network. M.S. Thesis, Xihua Normal University, Nanchong, China, 2020. [Google Scholar]
- Crouse, D.; Jacobs, R.L.; Richardson, Z.; Klum, S.; Jain, A.; Baden, A.L.; Tecot, S.R. LemurFaceID: A face recognition system to facilitate individual identification of lemurs. Bmc Zool. 2017, 2, 2. [Google Scholar] [CrossRef]
- Roy, A.M.; Bhaduri, J.; Kumar, T.; Raj, K. WilDect-YOLO: An efficient and robust computer vision-based accurate object localization model for automated endangered wildlife detection. Ecol. Inform. 2023, 75, 101919. [Google Scholar] [CrossRef]
- Akçay, H.G.; Kabasakal, B.; Aksu, D.; Demir, N.; Öz, M.; Erdoğan, A. Automated bird counting with deep learning for regional bird distribution mapping. Animals 2020, 10, 1207. [Google Scholar] [CrossRef]
- Dave, B.; Mori, M.; Bathani, A.; Goel, P. Wild Animal Detection using YOLOv8. Procedia Comput. Sci. 2023, 230, 100–111. [Google Scholar] [CrossRef]
- Schütz, A.K.; Schöler, V.; Krause, E.T.; Fischer, M.; Müller, T.; Freuling, C.M.; Lentz, H.H. Application of YOLOv4 for detection and Motion monitoring of red Foxes. Animals 2021, 11, 1723. [Google Scholar] [CrossRef]
- Xu, N.; Ma, Z.; Xia, Y.; Dong, Y.; Zi, J.; Xu, D.; Chen, F. A Serial Multi-Scale Feature Fusion and Enhancement Network for Amur Tiger Re-Identification. Animals 2024, 14, 1106. [Google Scholar] [CrossRef]
- Cronin, K.A. Prosocial behaviour in animals: The influence of social relationships, communication and rewards. Anim. Behav. 2012, 84, 1085–1093. [Google Scholar] [CrossRef]
- Vieira, M.; Fonseca, P.J.; Amorim, M.; Teixeira, C.J. Call recognition and individual identification of fish vocalizations based on automatic speech recognition: An example with the Lusitanian toadfish. J. Acoust. Soc. Am. 2015, 138, 3941–3950. [Google Scholar] [CrossRef]
- Elie, J.E.; Theunissen, F.E. Zebra finches identify individuals using vocal signatures unique to each call type. Nat. Commun. 2018, 9, 4026. [Google Scholar] [CrossRef]
- Verma, K.; Joshi, B. Different animal species hairs as biological tool for the forensic assessment of individual identification characteristics from animals of zoological park, Pragti Maidan, New Delhi, India. J. Forensic Res. 2012, 3, 2157–7145. [Google Scholar]
- Okura, F.; Ikuma, S.; Makihara, Y.; Muramatsu, D.; Nakada, K.; Yagi, Y. RGB-D video-based individual identification of dairy cows using gait and texture analyses. Comput. Electron. Agric. 2019, 165, 104944. [Google Scholar] [CrossRef]
- Rajkondawar, P.G.; Liu, M.; Dyer, R.M.; Neerchal, N.K.; Tasch, U.; Lefcourt, A.M.; Varner, M.A. Comparison of models to identify lame cows based on gait and lesion scores, and limb movement variables. J. Dairy Sci. 2006, 89, 4267–4275. [Google Scholar] [CrossRef]
- Farley, S.; Talbot, S.L.; Sage, G.K.; Sinnott, R.; Coltrane, J. Use of DNA from bite marks to determine species and individual animals that attack humans. Wildl. Soc. Bull. 2014, 38, 370–376. [Google Scholar] [CrossRef]
- Tillmar, A.O.; Dell’Amico, B.; Welander, J.; Holmlund, G. A universal method for species identification of mammals utilizing next generation sequencing for the analysis of DNA mixtures. PLoS ONE 2013, 8, e83761. [Google Scholar] [CrossRef]
- Kühl, H.S.; Burghardt, T. Animal biometrics: Quantifying and detecting phenotypic appearance. Trends Ecol. Evol. 2013, 28, 432–441. [Google Scholar] [CrossRef]
- Shi, C.; Liu, D.; Cui, Y.; Xie, J.; Roberts, N.J.; Jiang, G. Amur tiger stripes: Individual identification based on deep convolutional neural network. Integr. Zool. 2020, 15, 461–470. [Google Scholar] [CrossRef]
- Eikelboom, J.A.; Wind, J.; van de Ven, E.; Kenana, L.M.; Schroder, B.; de Knegt, H.J.; Prins, H.H. Improving the precision and accuracy of animal population estimates with aerial image object detection. Methods Ecol. Evol. 2019, 10, 1875–1887. [Google Scholar] [CrossRef]
- Perry, L.K.; Lupyan, G. Recognising a zebra from its stripes and the stripes from “zebra”: The role of verbal labels in selecting category relevant information. Lang. Cogn. Neurosci. 2017, 32, 925–943. [Google Scholar] [CrossRef]
- Lahiri, M.; Tantipathananandh, C.; Warungu, R.; Rubenstein, D.I.; Berger-Wolf, T.Y. Biometric animal databases from field photographs: Identification of individual zebra in the wild. In Proceedings of the 1st ACM International Conference on Multimedia Retrieval, Trento, Italy, 18–20 April 2011; pp. 1–8. [Google Scholar]
- Das, A.; Sowmya, S.; Sinha, S.; Chandru, S. Identification of a Zebra Based on Its Stripes through Pattern Recognition. In Proceedings of the 2021 International Conference on Design Innovations for 3Cs Compute Communicate Control (ICDI3C), Bangalore, India, 10–11 June 2021; pp. 120–122. [Google Scholar]
- Chen, G.; Han, T.X.; He, Z.; Kays, R.; Forrester, T. Deep convolutional neural network based species recognition for wild animal monitoring. In Proceedings of the 2014 IEEE International Conference on Image Processing (ICIP), Paris, France, 27–30 October 2014; pp. 858–862. [Google Scholar]
- Villa, A.G.; Salazar, A.; Vargas, F. Towards automatic wild animal monitoring: Identification of animal species in camera-trap images using very deep convolutional neural networks. Ecol. Inform. 2017, 41, 24–32. [Google Scholar] [CrossRef]
- Cheema, G.S.; Anand, S. Automatic detection and recognition of individuals in patterned species. In Proceedings of the Machine Learning and Knowledge Discovery in Databases: European Conference, ECML PKDD 2017, Skopje, Macedonia, 18–22 September 2017; Proceedings, Part III 10. Springer International Publishing: Cham, Switzerland, 2017; pp. 27–38. [Google Scholar]
- Norouzzadeh, M.S.; Nguyen, A.; Kosmala, M.; Swanson, A.; Palmer, M.S.; Packer, C.; Clune, J. Automatically identifying, counting, and describing wild animals in camera-trap images with deep learning. Proc. Natl. Acad. Sci. USA 2018, 115, E5716–E5725. [Google Scholar] [CrossRef]
- Gong, Y.; Tan, M.; Wang, Z.; Zhao, G.; Jiang, P.; Jiang, S.; Zhang, D.; Ge, K.; Feng, L. AI recognition of infrared camera image of wild animals based on deep learning: Northeast Tiger and Leopard National Park for example. Acta Theriol. Sin. 2019, 39, 458–465. [Google Scholar]
- Fan, Y. Design and Implementation of Golden Monkey Face Recognition Software based on Deep Learning. Master’s Thesis, Xidian University, Xi’an, China, 2018. [Google Scholar]
- Zhao, T.; Zhou, Z.; Li, D.; Liu, S.; Li, M. Individual Identification of Leopard Based on Improved Cifar-10 Deep Learning Model. J. Taiyuan Univ. Technol. 2018, 49, 585–591. [Google Scholar]
- Shi, C.M.; Xie, J.J.; Gu, J.Y.; Liu, D.; Jiang, G.S. Amur tiger individual automatic identification based on object detection. Acta Ecol. Sin. 2021, 41, 4685–4693. [Google Scholar]
- Li, S.; Li, J.; Tang, H.; Qian, R.; Lin, W. ATRW: A benchmark for Amur tiger re-identification in the wild. arXiv 2019, arXiv:1906.05586. [Google Scholar]
- Redmon, J.; Divvala, S.; Girshick, R.; Farhadi, A. You only look once: Unified, real-time object detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 26 June–1 July 2016; pp. 779–788. [Google Scholar]
- Szegedy, C.; Ioffe, S.; Vanhoucke, V.; Alemi, A. Inception-v4, inception-resnet and the impact of residual connections on learning. In Proceedings of the AAAI Conference on Artificial Intelligence, San Francisco, CA, USA, 4–9 February 2017; Volume 31. [Google Scholar]
- Srivastava, N.; Hinton, G.; Krizhevsky, A.; Sutskever, I.; Salakhutdinov, R. Dropout: A simple way to prevent neural networks from overfitting. J. Mach. Learn. Res. 2014, 15, 1929–1958. [Google Scholar]
- Woo, S.; Park, J.; Lee, J.Y.; Kweon, I.S. Cbam: Convolutional block attention module. In Proceedings of the European Conference on Computer Vision (ECCV), Munich, Germany, 8–14 September 2018; pp. 3–19. [Google Scholar]
- Hu, J.; Shen, L.; Sun, G. Squeeze-and-excitation networks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, 18–22 June 2018; pp. 7132–7141. [Google Scholar]
- Wang, Q.; Wu, B.; Zhu, P.; Li, P.; Zuo, W.; Hu, Q. ECA-Net: Efficient channel attention for deep convolutional neural networks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, WA, USA, 14–19 June 2020; pp. 11534–11542. [Google Scholar]
Model | P | R | [email protected] |
---|---|---|---|
Faster R-CNN | 56.23% | 94.87% | 82.73% |
SSD | 83.29% | 90.17% | 90.49% |
YOLOv5 | 88.90% | 95.00% | 97.30% |
Face | Left Stripe | Right Stripe | Complete Image | |
---|---|---|---|---|
Number of pictures | 1668 | 891 | 992 | 1887 |
Number of Amur tigers | 107 | 59 | 66 | 107 |
Hardware and Software Configuration | Versions |
---|---|
GPU | RTX 2080 Ti (11 GB) |
CPU | Intel(R) Xeon(R) Platinum 8255C CPU @ 2.50 GHz |
OS | Ubuntu |
CUDA | 11.2 |
TensorFlow | 2.5.0 |
Dropout Rate | Maximum Top1 Accuracy | Fifth Top1 Accuracy |
---|---|---|
0 | 0.9172 | 0.8828 |
0.1 | 0.9207 | 0.8897 |
0.2 | 0.9207 | 0.8793 |
0.3 | 0.9207 | 0.8828 |
0.4 | 0.9207 | 0.9000 |
0.5 | 0.9034 | 0.8690 |
Attention Mechanism | Accuracy Rate | Training Time/s |
---|---|---|
Original model | 0.9172 | 460 |
SE | 0.9034 | 474 |
ECA | 0.9138 | 472 |
CBAM | 0.9207 | 469 |
Layer (Type) | Output Shape | Param |
---|---|---|
inception_resnet_v2 (Functional) | (None, 8, 8, 1536) | 54,336,736 |
cbam_block (cbam_block) | (None, 8, 8, 1536) | 589,922 |
global_average_pooling2d_1 (GlobalAveragePooling2D) | (None, 1536) | 0 |
dropout (Dropout) | (None, 1536) | 0 |
dense_2 (Dense) | (None, 107) | 164,459 |
Model | Accuracy Rate | Training Time/s |
---|---|---|
VGG19 | 0.6517 | 104 |
ResNet50 | 0.8621 | 149 |
ResNet152 | 0.8414 | 340 |
InceptionV3 | 0.8862 | 194 |
InceptionResNetV2 | 0.9172 | 460 |
InceptionResNetV2 with CBAM | 0.9310 | 469 |
Model | Head | Left Body | Right Body |
---|---|---|---|
InceptionResNetV2 | 0.9172 | 0.9937 | 0.9186 |
InceptionResNetV2 with CBAM | 0.9310 | 0.9937 | 0.9360 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Wu, L.; Jinma, Y.; Wang, X.; Yang, F.; Xu, F.; Cui, X.; Sun, Q. Amur Tiger Individual Identification Based on the Improved InceptionResNetV2. Animals 2024, 14, 2312. https://doi.org/10.3390/ani14162312
Wu L, Jinma Y, Wang X, Yang F, Xu F, Cui X, Sun Q. Amur Tiger Individual Identification Based on the Improved InceptionResNetV2. Animals. 2024; 14(16):2312. https://doi.org/10.3390/ani14162312
Chicago/Turabian StyleWu, Ling, Yongyi Jinma, Xinyang Wang, Feng Yang, Fu Xu, Xiaohui Cui, and Qiao Sun. 2024. "Amur Tiger Individual Identification Based on the Improved InceptionResNetV2" Animals 14, no. 16: 2312. https://doi.org/10.3390/ani14162312