Quaternion Convolutional Neural Networks for Depth Estimation
An Hung Nguyen, Cao Duy Hoang, Dang Hoang Phu Phan, Minh Tuan Pham · 2022 Joint 12th International Conference on Soft Computing and Intelligent Systems and 23rd International Symposium on Advanced Intelligent Systems (SCIS&ISIS) · 2022
Depth Estimation is the task of measuring the distance of each pixel relative to the camera. Depth is extracted from either monocular (single) or stereo (multiple views of a scene) images. Traditional methods use multi-view geometry to find the relationship between the images [1]. Recently, Depth Estimation is interesting to global AI researchers. We can see applications of deep learning in solutions that. But deep learning methods before only research solutions in real-valued neural networks. In this paper, we want to propose a method, which is a quaternion-valued convolutional neural network because quaternion operations are great for depth processing and complex images, so neural networks that use quaternion-valued will have better than neural networks that use real-valued.