Reef fish identification using image recognition.

Big Game Fishing has been recognized as an emergent sport amongst the youth population of the Maldives and tourists today. As over a 1000 species of reef fish are inhabited in the country's vast sea, problems are faced within the youth to identify and inherit correct knowledge about the information of reef fish, which includes the species, names etc. This is also a challenge faced by the tourism sector, as the staff assisting the tourists in fishing trips have trouble providing specific important information of the fishes caught, which involves the species, Dhivehi, and English name of the fish, or whether it is a fish suitable for eating. The solution to this problem can be brought with the introduction of an application which recognizes reef fish using object detection algorithms and providing the available information of the reef fish identified, which would be informative for the target audience of this application. Emerging technologies, such as Artificial Intelligence (AI), specifically Object Detection Algorithms, is gaining popularity in assisting to detect and identify objects, hence making it also possible to identify the fish species using object detection and classification. To train the proposed model for identification of reef fish, the YOLOv7 model was used to generate the weights necessary for the detection algorithm. 5 classes (5 types of reef fish) were used, and for each class, 50 images were used for training the model and 10 images were used for validation respectively. This process has been run for 400 epochs and training and validation of the model was carried out in Google Colab. The computational speed of the detections had an average of 3.13 seconds, and the accuracy of detection was above 90%. [ABSTRACT FROM AUTHOR]