Attention Unet++ for lightweight depth estimation from sparse depth samples and a single RGB image

Tao, Zhao; Pan, Shuguo; Gao, Wang; Sheng, Chao; Sun, Yingchun; Wei, Jiansheng

doi:10.1007/s00371-021-02092-8

Cited by 13 publications

(3 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In a similar fashion, MAPUNet [58], inspired by UNet++ [66], and UNet 3+ [19], exploited multi-scale feature fusion and supervision for monocular depth estimation. Moreover, a UNet++ variant with residual blocks and dense gated convolution based attention [60] was used for monocular depth estimation using sparse depth measurements [64]. Finally, Na-sUNet [56] employed neural architecture search to look for an efficient and effective UNet architecture, a finding shared by [48] as well.…”

Section: Related Workmentioning

confidence: 99%

Hybrid Skip: A Biologically Inspired Skip Connection for the UNet Architecture

et al. 2022

View full text Add to dashboard Cite

In this work we introduce a biologically inspired long-range skip connection for the UNet architecture that relies on the perceptual illusion of hybrid images, being images that simultaneously encode two images. The fusion of early encoder features with deeper decoder ones allows UNet models to produce finer-grained dense predictions. While proven in segmentation tasks, the network's benefits are downweighted for dense regression tasks as these long-range skip connections additionally result in texture transfer artifacts. Specifically for depth estimation, this hurts smoothness and introduces false positive edges which are detrimental to the task due to the depth maps' piece-wise smooth nature. The proposed HybridSkip connections show improved performance in balancing the trade-off between edge preservation, and the minimization of texture transfer artifacts that hurt smoothness. This is achieved by the proper and balanced exchange of information that HybridSkip connections offer between the high and low frequency, encoder and decoder features, respectively. The code and models will be made available in the project page.

show abstract

Section: Related Workmentioning

confidence: 99%

Hybrid Skip: A Biologically Inspired Skip Connection for the UNet Architecture

et al. 2022

View full text Add to dashboard Cite

show abstract

“…In recent years, UNet++ neural network was proposed and later used by many scholars [23][24][25][26]. The UNet++ model is based on the UNet model and obtains the UNet++ network through nested connections.…”

Section: Introductionmentioning

confidence: 99%

Research on Road Extraction From High-Resolution Remote Sensing Images Based on Improved UNet++

Li,

Tan,

Xiao

et al. 2024

IEEE Access

View full text Add to dashboard Cite

To address the challenges of road extraction in high-resolution remote sensing images, this paper presents an enhanced UNet++ road extraction method that incorporates CBAM. The original UNet++ network is referenced, and the loss function is improved by introducing a new joint loss function. The enhanced UNet++ network utilizes an attention mechanism to enhance the network's ability to identify road features, thereby improving the accuracy of road extraction. Additionally, a new joint loss function is employed to enhance the network's stability and further improve its road extraction capability. Experimental validation is performed on the Massachusetts roads dataset and DeepGlobal road dataset. The experimental results demonstrate that this method outperforms U-Net, SegNet, and UNet++ networks in terms of IoU, Recall, OA, and Kappa. Specifically, on the Massachusetts road dataset, the OA and Kappa values are 94.92% and 0.9202, respectively. On the DeepGlobal road dataset, the OA and Kappa values for this algorithm are 98.12% and 0.9515, respectively. The ablation experiment confirms the effectiveness of the proposed enhancements. In conclusion, this paper presents a method that effectively extracts roads from high-resolution remote sensing images, exhibits a certain level of generalization ability, and can provide valuable support for road protection and planning.

show abstract

“…To overcome this limitation, lightweight networks for depth completion tasks were proposed. Tao et al introduced lightweight depth completion with a Sobel edge prediction network [ 11 ] and self-attention-based multi-level feature integration and extraction [ 12 ]. Although these approaches contribute to decreasing the computational cost by effectively reducing the parameter size and model complexity, they cannot leverage or surpass the better performance of existing networks.…”

Section: Introductionmentioning

confidence: 99%

Lightweight Depth Completion Network with Local Similarity-Preserving Knowledge Distillation

Jeong

Park

Cho

et al. 2022

Sensors

View full text Add to dashboard Cite

Depth perception capability is one of the essential requirements for various autonomous driving platforms. However, accurate depth estimation in a real-world setting is still a challenging problem due to high computational costs. In this paper, we propose a lightweight depth completion network for depth perception in real-world environments. To effectively transfer a teacher’s knowledge, useful for the depth completion, we introduce local similarity-preserving knowledge distillation (LSPKD), which allows similarities between local neighbors to be transferred during the distillation. With our LSPKD, a lightweight student network is precisely guided by a heavy teacher network, regardless of the density of the ground-truth data. Experimental results demonstrate that our method is effective to reduce computational costs during both training and inference stages while achieving superior performance over other lightweight networks.

show abstract

Attention Unet++ for lightweight depth estimation from sparse depth samples and a single RGB image

Cited by 13 publications

References 41 publications

Hybrid Skip: A Biologically Inspired Skip Connection for the UNet Architecture

Hybrid Skip: A Biologically Inspired Skip Connection for the UNet Architecture

Research on Road Extraction From High-Resolution Remote Sensing Images Based on Improved UNet++

Lightweight Depth Completion Network with Local Similarity-Preserving Knowledge Distillation

Contact Info

Product

Resources

About