URCDC-Depth: Uncertainty Rectified Cross-Distillation with CutFlip for Monocular Depth Estimation

This work aims to estimate a high-quality depth map from a single RGB image. Due to the lack of depth clues, making full use of the long-range correlation and the local information is critical for accurate depth estimation. Towards this end, we introduce an uncertainty rectified cross-distillation between Transformer and convolutional neural network (CNN) to learn a unified depth estimator. Specifically, we use the depth estimates derived from the Transformer branch and the CNN branch as pseudo labels to teach each other. Meanwhile, we model the pixel-wise depth uncertainty to rectify the loss weights of noisy depth labels. To avoid the large performance gap induced by the strong Transformer branch deteriorating the cross-distillation, we transfer the feature maps from Transformer to CNN and design coupling units to assist the weak CNN branch to utilize the transferred features. Furthermore, we propose a surprisingly simple yet highly effective data augmentation technique CutFlip, which enforces the model to exploit more valuable clues apart from the clue of vertical image position for depth estimation. Extensive experiments indicate that our model, termed~\textbf{URCDC-Depth}, exceeds previous state-of-the-art methods on the KITTI and NYU-Depth-v2 datasets, even with no additional computational burden at inference time. The source code is publicly available at \url{https://github.com/ShuweiShao/URCDC-Depth}.