Siralama Sorunu Olarak Nispi Derinlik Tahmini Relative Depth Estimation as a Ranking Problem

Abstract

Özetçe -Bu bildiride, tekil görüntüden nispi derinlik tahmin etme problemini, sıralama problemi olarak ele alan ve çözüm sunan yeni bir yöntem sunuyoruz. Derinlik tahmini problemini bu şekilde yeniden formüle ederek, sıralama problemi üzerine var olan literatürden faydalanabildik ve mevcut bilgileri daha iyi sonuçlar elde etmek için uygulayabildik. Bu amaçla, sıralama literatüründen ödünç aldıgımız agırlıklandırılmış ListMLE kayıp fonksiyonunu nispi derinlik tahmini problemine kazandırdık. Ayrıca, yöntemimizin daha güçlü oldugu piksel derinligi sıralama dogrulugunu dikkate alan yeni bir metrik getirdik.Anahtar Kelimeler-derin ögrenme, derinlik tahmini, nispi derinlik tahmini, sıralamayı ögrenme.

Single Image Depth Estimation: An Overview

“…An important theme we see here is: Classification and ordinality A number of works decided to formulate the depth estimation problem in a way that does not require the system to estimate the exact depth value [2], [12], [20], [27], [40], [41], [42], [43], [44], [45]. While estimating the relative depth instead of the absolute depth is an alternative, estimating the depth range in a classification setting can be done with success.…”

“…There are two important issues to consider for depth in the wild problem. The first issue a system faces that aims to work in [20] softmax-loss -custom Chen et al [12] pairwise ranking -Hourglass Xian et al [40] pairwise ranking random + hard pairs EncDecResNet Chen et al [41] pairwise ranking random + hard pairs EncDecResNet Xian et al [44] pairwise ranking random + instance edges + image edges EncDecResNet Mertan et al [42] listwise ranking random EncDecResNet Lienen et al [45] listwise ranking random EfficientNet variant Mertan et al [43] distributional ranking random EncDecResNet the wild is that the depth ranges may vary quite a lot which is a problem for the learning process. While the range of absolute depth values for an image of a table may be between 10 cm and a couple of meters, the image of a touristic place such as "Eiffel Tower" contains pixels that are a hundred meters away, if they have a depth value at all (sky regions can be considered infinitely far away).…”

“…Instead of improving the sampling strategy, Mertan et al [42] and Lienen et al [45] try to improve the loss function. Both work implements the same idea.…”

“…They adopt a listwise ranking loss function, the ListMLE, that considers a list of pixels per training instance, as opposed to a pair of pixels. Additionally, [42] advocates the usage of the mean average precision (MAP) as a metric to measure the performance in the test time, instead of accuracy, since MAP is capable of emphasizing the correct ranking of closer pixels, which is arguably more important for the depth estimation problem.…”