Evaluation Of The Performance Of Distinctive Object Detection And Camouflaged Object Detection Models Using Night Image Data

Kılıç A., Tümüklü Özyer G., Göktaş Y., Usanmaz B.

4th Eurasian Conference on Human-Computer Interaction, Baku, Azerbaycan, 5 - 06 Aralık 2025, ss.33-37, (Tam Metin Bildiri)

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Basıldığı Şehir: Baku
  • Basıldığı Ülke: Azerbaycan
  • Sayfa Sayıları: ss.33-37
  • Atatürk Üniversitesi Adresli: Evet

Özet

Object detection in nighttime images is a challenge that traditional object detection models struggle to solve due to insufficient light and ambiguities. Therefore, specific models for this domain are being designed and proposed in the literature. Numerous studies have demonstrated that traditional models are inadequate and unsuitable for this domain, but there are very few studies examining the performance of models developed for more specific tasks, such as distinct object detection (SOD) or camouflaged object detection (COD), and whether they can contribute to this field. In this study, different SOD and COD models were trained on a dataset containing nighttime images, and the results were evaluated. The results showed that camouflaged object detection models outperformed other models.