30 August 2023 in IEEE Transactions on Multimedia
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 local information is critical for accurate depth estimation. To this end, we introduce an uncertainty rectified cross-distillation between the Transformer and convolutional neural network (CNN) to achieve a comprehensive depth estimator. Specifically, we utilize the depth estimates from the Transformer branch and CNN branch as pseudo labels to teach each other. At the same time, the pixel-wise depth uncertainty is modeled to mitigate the negative impact of noisy pseudo labels. To avoid the large capacity gap induced by the strong Transformer branch deteriorating the cross-distillation, we transfer the feature maps from the Transformer to the CNN and develop coupling units to assist the weak CNN branch in leveraging the transferred features. Furthermore, we introduce CutFlip, a surprisingly simple yet highly effective data augmentation technique, which forces the model to focus on more valuable depth reasoning clues apart from the vertical image position. Extensive experiments demonstrate that our model, termed URCDC-Depth , exceeds in performance previous state-of-the-art approaches on the KITTI, NYU-Depth-v2 and SUN RGB-D datasets, with no additional computational burden in the evaluation phase. The source code will be publicly available upon acceptance.
12 May 2023 in IEEE Transactions on Circuits and Systems for Video Technology
Monocular depth estimation plays a fundamental role in computer vision. Due to the costly acquisition of depth ground truth, self-supervised methods that leverage adjacent frames to establish a supervision signal have emerged as the most promising paradigms. In this work, we propose two novel ideas to improve self-supervised monocular depth estimation: 1) self-reference distillation and 2) disparity offset refinement. Specifically, we use a parameter-optimized model as the teacher updated as the training epochs to provide additional supervision during the training process. The teacher model has the same structure as the student model, with weights inherited from the historical student model. In addition, a multiview check is introduced to filter out the outliers produced by the teacher model. Furthermore, we leverage the contextual consistency between high-level and low-level features to obtain multiscale disparity offsets, which are used to refine the disparity output incrementally by aligning disparity information at different scales. The experimental results on the KITTI and Make3D datasets show that our method outperforms previous state-of-the-art competitors.
2 December 2024 in IEEE Transactions on Circuits and Systems for Video Technology
Over the past few years, self-supervised monocular depth estimation has received widespread attention. Most efforts focus on designing different types of network architectures and loss functions or handling edge cases, for example, occlusion and dynamic objects. In this work, we take another path and propose a novel conditional diffusion-based generative framework for self-supervised monocular depth estimation, dubbed MonoDiffusion. Because the depth ground-truth is unavailable in a self-supervised setting, we develop a new pseudo ground-truth diffusion process to assist the diffusion for training. Instead of diffusing at a fixed high resolution, we perform diffusion in a coarse-to-fine manner that allows for faster inference time without sacrificing accuracy or even better accuracy. Furthermore, we develop a simple yet effective contrastive depth reconstruction mechanism to enhance the denoising ability of model. It is worth noting that the proposed MonoDiffusion has the property of naturally acquiring the depth uncertainty that is essential to be implemented in safety-critical cases. Extensive experiments on the KITTI, Make3D and DIML datasets indicate that our MonoDiffusion outperforms prior state-of-the-art self-supervised competitors. The source code will be publicly available upon the acceptance.
10 June 2024 in IEEE Transactions on Pattern Analysis and Machine Intelligence
Over the past few years, monocular depth estimation and completion have been paid more and more attention from the computer vision community because of their widespread applications. In this paper, we introduce novel physics (geometry)-driven deep learning frameworks for these two tasks by assuming that 3D scenes are constituted with piece-wise planes. Instead of directly estimating the depth map or completing the sparse depth map, we propose to estimate the surface normal and plane-to-origin distance maps or complete the sparse surface normal and distance maps as intermediate outputs. To this end, we develop a normal-distance head that outputs pixel-level surface normal and distance. Afterthat, the surface normal and distance maps are regularized by a developed plane-aware consistency constraint, which are then transformed into depth maps. Furthermore, we integrate an additional depth head to strengthen the robustness of the proposed frameworks. Extensive experiments on the NYU-Depth-v2, KITTI and SUN RGB-D datasets demonstrate that our method exceeds in performance prior state-of-the-art monocular depth estimation and completion competitors.
25 November 2022 in IEEE Transactions on Multimedia
Depth estimation attracts widespread attention in the computer vision community. However, it is still quite difficult to recover an accurate depth map using only one RGB image. We observe a phenomenon that existing methods tend to fail in different cases, caused by differences in network architecture, loss function and so on. In this work, we investigate into the phenomenon and propose to integrate the strengths of multiple weak depth predictor to build a comprehensive and accurate depth predictor, which is critical for many real-world applications, e.g. , 3D reconstruction. Specifically, we construct multiple base (weak) depth predictors by utilizing different Transformer-based and convolutional neural network (CNN)-based architectures. Transformer establishes long-range correlation while CNN preserves local information ignored by Transformer due to the spatial inductive bias. Therefore, the coupling of Transformer and CNN contributes to the generation of complementary depth estimates, which are essential to achieve a comprehensive depth predictor. Then, we design mixers to learn from multiple weak predictions and adaptively fuse them into a strong depth estimate. The resultant model, which we refer to as Transformer-assisted depth ensembles (TEDepth). On the standard NYU-Depth-v2 and KITTI datasets, we thoroughly explore how the neural ensembles affect the depth estimation and demonstrate that our TEDepth achieves better results than previous state-of-the-art approaches. To validate the generalizability across cameras, we directly apply the models trained on NYU-Depth-v2 to the SUN RGB-D dataset without any fine-tuning, and the superior results emphasize its strong generalizability. The source code and trained models will be publicly available upon the acceptance.
31 July 2023 in IEEE Transactions on Intelligent Vehicles
Monocular depth estimation has attracted extensive attention and made great progress in recent years. However, the performance still lags far behind LiDAR-based depth completion algorithms. This is because the completion algorithms not only utilize the RGB image, but also have the prior of sparse depth collected by LiDAR. To reduce this performance gap, we propose a novel initiative that incorporates the concept of pseudo-LiDAR into depth estimation. The pseudo-LiDAR depends only on the camera and thus achieves a lower cost than LiDAR. To emulate the scan pattern of LiDAR, geometric sampling and appearance sampling are proposed. The former measures the vertical and horizontal azimuths of 3D scene points to establish the geometric correlation. The latter helps determine which “ pseudo-LiDAR rays” return an answer and which do not. Then, we build a sparse pseudo-LiDAR-based depth estimation framework. Extensive experiments show that the proposed method surpasses previous state-of-the-art competitors on the KITTI, NYU-Depth-v2 and SUN RGB-D datasets. The source code and trained models will be publicly available upon acceptance.