Questions about training data preparation

nobukatsu-kajiura / UR-COD

Official implementation for the paper "Improving Camouflaged Object Detection with the Uncertainty of Pseudo-edge Labels"

3 stars 1 forks source link

Questions about training data preparation #1

Closed jokerhcy closed 2 years ago

jokerhcy commented 3 years ago

Thanks to the author for sharing in COD tasks. I encountered a detailed problem in the process of Paper recurrence: What do Images_Gray specifically refer to in ./data/TrainDataset/Img_Gray/ for the preparation of training? It has not been mentioned in this papers.

nobukatsu-kajiura commented 2 years ago

Thanks for your question. This implementation refers to UCNet[1] and the smoothness loss used in this model requires a gray-scale image. [1] Zhang et al. UC-Net: Uncertainty Inspired RGB-D Saliency Detection via Conditional Variational Autoencoders. In CVPR. 2020.