1. Home
  2. Browse by Author

Browsing by Author "Can, Duygu"

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • Results Per Page
  • Sort Options
  • Loading...
    Thumbnail Image
    master's-degree-project.listelement.badge
    Humpback Whale Indentification With Convolutional Neural Networks
    (MEF Üniversitesi, Fen Bilimleri Enstitüsü, 2018) Can, Duygu; Arslan, Şuayb Ş.
    The migration patterns of humpback whales are tracked with conventional photoidentification techniques for decades. The distinct markings on whale flukes serve as unique fingerprints for these creatures. This study aims to identify humpback whales according to their fluke images using ResNET, a deep neural network architecture to help the conservation efforts for this endangered species by automatizing the process. We experimented with different train/test split schemes and initializations to obtain the best classifying model. Although we were limited with a small sized training set of 200 images, using state-of-the-art image processing and data augmentation methods we obtained a high accuracy of 0.94 for 11 distinct whales. This project is served as an friendly interface to dive deep into the field of image recognition with Convolutional Neural Networks.