Abstract
Fishermen need to remove abnormal or dead fish for the prevention of viral infection. However, the identification of diseased fish is more ambiguous that the identification of dead ones, and detecting abnormal fish is hard even for experienced fishermen to do. Therefore, the automatic detection of abnormal fish is desirable. However, the detection of abnormal fish by Deep Learning needs a lot of image data both of healthy fish and abnormal ones. There are many image data of healthy fish, but image data of abnormal fish are scarce. In situations where a large amount of data is not available, Fine-Tuning is commonly used. However, preprocessing is important in Fine-Tuning. In this research, we verify appropriate preprocessing for detection of abnormal fish by Fine-Tuning. Appropriate preprocessing for Fine-Tuning using VGG16, which is used for image recognition, resulted in an increase in Recall by 12.5 points compared to conventional methods.
Similar content being viewed by others
References
Ishiguro Y, Gotoh M, Akinaka K (2003) Development of transmission system for underwater video camera picture and observation of fish schools in set-net. Bull Kanagawa Prefect Fish Res Inst 8:39–45
Yamashita Y (2020) Confirmation of upstream fish movement along the fish passage with stacking boulders through underwater imaging using time-lapse camera. Bulletin of the okayama prefectural technology center for agriculture, forestry, and fisheries research institute for fisheries science 35:33–37
Nakayama H (2015) Image feature extraction and transfer learning using deep convolutional neural networks. IEICE Tech Report 115(146):55–59
Iyatomi H (2019) Trends and challenges of automatic diagnosis techniques for plant diseases. Brain Neural Netw 26(4):123–134
Kawahara R, Nobuhara S, Matsuyama T (2016) Dynamic 3D capture of swimming fish by underwater active stereo. Methods Oceanogr 17:118–137
Heckbert P (1994) GRAPHICS GEMS IV, pp 474–485, Morgan Kaufmann, place
Pradeep Kumar Reddy R, Nagaraju C, Rajasekhar Reddy I (2015) Canny scale edge detection. IJETT. https://doi.org/10.14445/22315381/IJETT-ICGTETM-N3/ICGTETM-P121
Radenovic F, Tolias G, Chum O (2019) Fine-tuning CNN image retrieval with no human annotation. IEEE Trans Pattern Anal Mach Intell 41(7):1655–1668
Miyazaki S, Ooto T, Iwahori Y, Boonserm K (2018) Automatic polyp detection from endoscope image using transfer learning. Proc Fuzzy Syst Sympos 34:796–800
Simonyan K and Zisserman A (2015) Very deep convolutional networks for large-scale image recognition. ICLR. https://doi.org/10.48550/arXiv.1409.1556
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This work was presented in part at the joint symposium of the 27th International Symposium on Artificial Life and Robotics, the 7th International Symposium on BioComplexity, and the 5th International Symposium on Swarm Behavior and Bio-Inspired Robotics (Online, January 25–27, 2022).
About this article
Cite this article
Okawa, R., Iwasaki, N., Okamoto, K. et al. Detection of abnormal fish by image recognition using fine-tuning. Artif Life Robotics 28, 175–180 (2023). https://doi.org/10.1007/s10015-022-00824-0
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10015-022-00824-0