Challenges in Determining the Depth in 2-D Images

Akin Z., Sayar A.

16th International Conference on INnovations in Intelligent SysTems and Applications, INISTA 2022, Biarritz, Fransa, 8 - 12 Ağustos 2022 identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/inista55318.2022.9894120
  • Basıldığı Şehir: Biarritz
  • Basıldığı Ülke: Fransa
  • Anahtar Kelimeler: 3D bounding box, deep learning, depth estimation, homography, image processing, perspective
  • Kocaeli Üniversitesi Adresli: Evet

Özet

© 2022 IEEE.When an object is pictured, it's projection is changed from 3-D scene to 2-D plane. In these cases, the depth information of the object is lost. Depth allows the calculation of an object's distance and its three-dimensional perception. Using two or even more 2-D pictures from various perspectives of same scene aids in estimating relative depth information. However, if only one 2-D image is used, then the depth estimation problem becomes very hard to solve. In this study, we present the theoretical basis of this problem, summarize the proposed solution approaches, and examines the challenges.