RT journal article
T1 Enhancing Fish Auction with Deep Learning and Computer Vision: Automated Caliber and Species Classification
A1 Jareño Dorado, Javier
A1 Barcena González, Guillermo
A1 Castro-Gutiérrez, Jairo
A1 Cabrera Castro, Remedios
A1 Galindo Riaño, Pedro Luis
A2 Biología
A2 Ingeniería Informática
K1 fish species
K1 fish size
K1 machine learning
K1 deep learning
K1 transfer learning
K1 fish auction
K1 classification
AB The accurate labeling of species and size of specimens plays a pivotal role in fish auctionsconducted at fishing ports. These labels, among other relevant information, serve as determinantsof the objectivity of the auction preparation process, underscoring the indispensable nature of areliable labeling system. Historically, this task has relied on manual processes, rendering it vulnerableto subjective interpretations by the involved personnel, therefore compromising the value of themerchandise. Consequently, the digitization and implementation of an automated labeling systemare proposed as a viable solution to this ongoing challenge. This study presents an automatic systemfor labeling species and size, leveraging pre-trained convolutional neural networks. Specifically,the performance of VGG16, EfficientNetV2L, Xception, and ResNet152V2 networks is thoroughlyexamined, incorporating data augmentation techniques and fine-tuning strategies. The experimentalfindings demonstrate that for species classification, the EfficientNetV2L network excels as the mostproficient model, achieving an average F-Score of 0.932 in its automatic mode and an averageF-Score of 0.976 in its semi-automatic mode. Concerning size classification, a semi-automatic modelis introduced, where the Xception network emerges as the superior model, achieving an averageF-Score of 0.949.
SN 2410-3888
YR 2024
FD 2024-04-13
LK http://hdl.handle.net/10498/31770
UL http://hdl.handle.net/10498/31770
LA eng
DS Repositorio Institucional de la Universidad de Cádiz
RD 08-nov-2026