Browsing by Author "Ventura, Renz Michael D."

Now showing 1 - 1 of 1
  • No Thumbnail Available
    Publication
    FISH SCAN: a mobile application to prevent intermixing of Ladyfish (Elops hawaienis) fingerling in Milkfish (Chanos chanos) pond cultivation
    (Don Mariano Marcos Memorial State University - South La Union Campus, 2024-05) Lachica, Willy Boy S.; Balangue, Lee Jelyn A.; Galera, Mark T.; Gali, Jessa Mea V.; Ventura, Renz Michael D.
    As ladyfish are pests in milkfish pond cultivation, their cannibalistic behavior towards milkfish .fingerlings significantly hampers the growth rate of the cultivated milkfish. The similar physical appearance of ladyfish often deceives cultivators, leading them to mistakenly identify these pests as milkfish further complicating the cultivation process. To address this issue, the researchers propose "Fish-Scan, " a mobile application designed to detect ladyfish using object detection. The application employs the YOLOv8 architecture for real-time object detection and good accuracy. Four major experiments were conducted to train and test the model, utilizing metrics such as Fl-score and mAP50 to determine the best-performing model. The model demonstrating the highest performance featured a white background with YOLOv8 default augmentation, achieving an F i-score of 0.98 and a mAP50 of 0.99. Despite the model's training under 200 epochs, it achieved early convergence. The model was exported to ONNX runtime for deployment and integrated into Android Studio for application development. While real-time performance remains a challenge, the model has good performance in detecting captured images.