Deep Learning based Individual Cattle Face Recognition using Data Augmentation and Transfer Learning

POLAT H. E., Koc D. G., Ertugrul O., KOÇ C., Ekinci K.

JOURNAL OF AGRICULTURAL SCIENCES-TARIM BILIMLERI DERGISI, cilt.31, sa.1, ss.137-150, 2025 (SCI-Expanded, Scopus, TRDizin)

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 31 Sayı: 1
  • Basım Tarihi: 2025
  • Doi Numarası: 10.15832/ankutbd.1509798
  • Dergi Adı: JOURNAL OF AGRICULTURAL SCIENCES-TARIM BILIMLERI DERGISI
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, CAB Abstracts, Food Science & Technology Abstracts, Veterinary Science Database, TR DİZİN (ULAKBİM)
  • Sayfa Sayıları: ss.137-150
  • Ankara Üniversitesi Adresli: Evet

Özet

Accurate identification of cattle is essential for monitoring ownership, controlling production supply, preventing disease, and ensuring animal welfare. Despite the widespread use of ear tag-based techniques in livestock farm management, large-scale farms encounter challenges in identifying individual cattle. The process of identifying individual animals can be hindered by ear tags that fall off, and the ability to identify them over a long period of time becomes impossible when tags are missing. A dataset was generated by capturing images of cattle in their native environment to tackle this issue. The dataset was divided into three segments: training, validation, and testing. The dataset consisted of 15 000 records, each pertaining to a distinct bovine specimen from a total of 30 different cattle. To identify specific cattle faces in this study, deep learning algorithms such as InceptionResNetV2, MobileNetV2, DenseNet201, Xception, and NasNetLarge were utilized. The DenseNet201 algorithm attained a peak test accuracy of 99.53% and a validation accuracy of 99.83%. Additionally, this study introduces a novel approach that integrates advanced image processing techniques with deep learning, providing a robust framework that can potentially be applied to other domains of animal identification, thus enhancing overall farm management and biosecurity.