Abstract
Guided by information from RGB images, depth completion methods rebuild the dense depth from sparse depth input. However, the varying densities of valid pixels in sparse depth maps pose a significant challenge to the robustness of the completion model. To improve the robustness of depth completion, we propose a two-stage model called Semantic Aggregated Depth Completion (SADC) in this paper, comprising a coarse-grained completion stage and a fine-grained completion stage. In the coarse-grained completion stage, the Semantic Extraction Network (SEN) extracts RGB features and sends them to the Dynamic Semantic Aggregation (DSA) to predict the local semantic relationship (LSR) matrix. DSA aggregates the valid information based on the LSR matrix iteratively, resulting in coarse-grained completion results. In the fine-grained completion stage, SADC uses the Semantic Guidance Network (SGN) and Semantic Guidance Fusion (SGF) modules to refine the dense depth features from coarse-grained completion results by RGB features in multi-level and predict fine-grained completion results. We validate our method on NYU-v2 and KITTI with different valid pixel densities. The results demonstrate that SADC performs best results on benchmark tests and exhibits robustness to different densities without retraining.
Graphical abstract
Similar content being viewed by others
Availability of data and materials
NYU-v2 Benchmark: https://cs.nyu.edu/~silberman/datasets/nyu_depth_v2.html KITTI Benchmark: https://www.cvlibs.net/datasets/kitti
References
Song Z, Lu J, Yao Y et al (2021) Self-supervised depth completion from direct visual-lidar odometry in autonomous driving. IEEE Trans Intell Transp Syst 23(8):11654–11665. https://doi.org/10.1109/TITS.2021.3106055
Dai X, Yuan X, Wei X (2020) Tirnet: object detection in thermal infrared images for autonomous driving. Appl Intell 51:1244–1261. https://doi.org/10.1007/s10489-020-01882-2
Liu S, Bai X, Fang M et al (2021) Mixed graph convolution and residual transformation network for skeleton-based action recognition. Appl Intell 52:1544–1555. https://doi.org/10.1007/s10489-021-02517-w
Shukla P, Pramanik N, Mehta D et al (2022) Generative model based robotic grasp pose prediction with limited dataset. Appl Intell 52:9952–9966. https://doi.org/10.1007/s10489-021-03011-z
Tan Z, Gu X (2021) Depth scale balance saliency detection with connective feature pyramid and edge guidance. Appl Intell 51:5775–5792. https://doi.org/10.1007/s10489-020-02150-z
Luo H, Gao Y, Wu Y et al (2019) Real-time dense monocular SLAM with online adapted depth prediction network. IEEE Trans Multimed 21(2):470–483. https://doi.org/10.1109/TMM.2018.2859034
Sun J, Wang Z, Yu H et al (2022) Two-stage deep regression enhanced depth estimation from a single RGB image. IEEE Trans Emerg Top Comput 10(2):719–727. https://doi.org/10.1109/TETC.2020.3034559
Zhao ZQ, Zheng P, St Xu et al (2019) Object detection with deep learning: a review. IEEE Trans Neur Net Lear 30(11):3212–3232. https://doi.org/10.1109/TNNLS.2018.2876865
Pang Y, Li Y, Shen J et al (2019) Towards bridging semantic gap to improve semantic segmentation. In: International conference on computer vision (ICCV). IEEE, pp 4229–4238. https://doi.org/10.1109/ICCV.2019.00433
Crespo J, Castillo JC, Mozos ÓM et al (2020) Semantic information for robot navigation: a survey. Appl Sci. https://doi.org/10.3390/app10020497
Hu J, Ozay M, Zhang Y et al (2019) Revisiting single image depth estimation: toward higher resolution maps with accurate object boundaries. In: IEEE winter conference on applications of computer vision (WACV). IEEE, pp 1043–1051. https://doi.org/10.1109/WACV.2019.00116
Bhat SF, Alhashim I, Wonka P (2021) Adabins: depth estimation using adaptive bins. In: Conference on computer vision and pattern recognition (CVPR). Computer Vision Foundation / IEEE, p 4009–4018. https://doi.org/10.1109/CVPR46437.2021.00400
Piccinelli L, Sakaridis C, Yu F (2023) idisc: internal discretization for monocular depth estimation. 2023 IEEE/CVF conference on computer vision and pattern recognition (CVPR), pp 21477–21487. https://doi.org/10.1109/CVPR52729.2023.02057
Wang T, Ray N (2023) Compact depth-wise separable precise network for depth completion. IEEE Access 11:72679–72688. https://doi.org/10.1109/ACCESS.2023.3294247
Li Y, Jung C (2023) Deep sparse depth completion using joint depth and normal estimation. In: IEEE international symposium on circuits and systems. IEEE, pp 1–5. https://doi.org/10.1109/ISCAS46773.2023.10181618
Liu L, Song X, Lyu X et al (2021) Fcfr-net: feature fusion based coarse-to-fine residual learning for depth completion. In: The AAAI conference on artificial intelligence, pp 2136–2144. https://doi.org/10.1609/aaai.v35i3.16311
Wang H, Wang M, Che Z et al (2022) Rgb-depth fusion GAN for indoor depth completion. In: Conference on computer vision and pattern recognition (CVPR). IEEE, pp 6199–6208. https://doi.org/10.1109/CVPR52688.2022.00611
Tang J, Tian F, Feng W et al (2020) Learning guided convolutional network for depth completion. IEEE Trans Image Process 30:1116–1129. https://doi.org/10.1109/TIP.2020.3040528
Wang Y, Li B, Zhang G et al (2023) LRRU: long-short range recurrent updating networks for depth completion. In: International conference on computer vision (ICCV). IEEE, pp 9388–9398. https://doi.org/10.1109/ICCV51070.2023.00864
Palmer Stephen E (1999) Vision science: photons to phenomenology. Q Rev Biol 77(4):233–234. https://doi.org/10.1086/420636
Karsch K, Liu C, Kang SB (2014) Depth transfer: depth extraction from video using non-parametric sampling. IEEE Trans Pattern Anal Mach Intell 36(11):2144–2158. https://doi.org/10.1109/TPAMI.2014.2316835
Konrad J, Wang M, Ishwar P (2012) 2d-to-3d image conversion by learning depth from examples. In: Proceedings of IEEE conference on computer vision and pattern recognition workshops. IEEE Computer Society, pp 16–22. https://doi.org/10.1109/CVPRW.2012.6238903
Saxena A, Chung S, Ng A (2005) Learning depth from single monocular images. In: Advances in neural information processing systems, pp 1161–1168
Hoiem D, Efros AA, Hebert M (2005) Automatic photo pop-up. In: International conference and exhibition on computer graphics and interactive techniques, pp 577–584. https://doi.org/10.1145/1073204.1073232
Suwajanakorn S, Hernandez C, Seitz SM (2015) Depth from focus with your mobile phone. In: Conference on computer vision and pattern recognition (CVPR). IEEE Computer Society, pp 3497–3506. https://doi.org/10.1109/CVPR.2015.7298972
Yang X, Chang Q, Liu X et al (2021) Monocular depth estimation based on multi-scale depth map fusion. IEEE Access 9:67696–67705. https://doi.org/10.1109/ACCESS.2021.3076346
Zhang A, Ma Y, Liu J et al (2023) Promoting monocular depth estimation by multi-scale residual laplacian pyramid fusion. IEEE Signal Process Lett 30:205–209. https://doi.org/10.1109/LSP.2023.3251921
Wang L, Zhang J, Wang O et al (2020) Sdc-depth: semantic divide-and-conquer network for monocular depth estimation. In: Conference on computer vision and pattern recognition (CVPR). Computer Vision Foundation / IEEE, pp 538–547. https://doi.org/10.1109/CVPR42600.2020.00062
Liu Y (2020) Multi-scale spatio-temporal feature extraction and depth estimation from sequences by ordinal classification. Sensors (Basel, Switzerland) 20. https://doi.org/10.3390/s20071979
Meng X, Fan C, Ming Y et al (2022) Cornet: context-based ordinal regression network for monocular depth estimation. IEEE Trans Circuits Syst Video Technol 32:4841–4853. https://doi.org/10.1109/TCSVT.2021.3128505
Lee JH, Kim CS (2019) Monocular depth estimation using relative depth maps. In: Conference on computer vision and pattern recognition (CVPR). Computer Vision Foundation / IEEE, pp 9729–9738. https://doi.org/10.1109/CVPR.2019.00996
Lee JH, Kim CS (2022) Single-image depth estimation using relative depths. J Vis Commun Image Represent 84:103459. https://doi.org/10.1016/j.jvcir.2022.103459
Uhrig J, Schneider N, Schneider L et al (2017) Sparsity invariant cnns. In: International conference on 3D vision (3DV). IEEE Computer Society, pp 11–20. https://doi.org/10.1109/3DV.2017.00012
Knutsson H, Westin CF (1993) Normalized and differential convolution. In: Conference on computer vision and pattern recognition (CVPR). IEEE, pp 515–523. https://doi.org/10.1109/CVPR.1993.341081
Eldesokey A, Felsberg M, Khan FS (2019) Confidence propagation through cnns for guided sparse depth regression. IEEE Trans Pattern Anal Mach Intell 42(10):2423–2436. https://doi.org/10.1109/TPAMI.2019.2929170
Eldesokey A, Felsberg M, Holmquist K et al (2020) Uncertainty-aware cnns for depth completion: uncertainty from beginning to end. In: Conference on computer vision and pattern recognition (CVPR). Computer Vision Foundation / IEEE, pp 12011–12020. https://doi.org/10.1109/CVPR42600.2020.01203
Tomasi C, Manduchi R (1998) Bilateral filtering for gray and color images. In: International conference on computer vision (ICCV). IEEE Computer Society, pp 839–846. https://doi.org/10.1109/ICCV.1998.710815
Levin A, Lischinski D, Weiss Y (2004) Colorization using optimization. In: International conference and exhibition on computer graphics and interactive techniques, pp 689–694. https://doi.org/10.1145/1015706.1015780
Qu C, Nguyen T, Taylor CJ (2020) Depth completion via deep basis fitting. IEEE Winter conference on applications of computer vision (WACV), pp 71–80. https://doi.org/10.1109/WACV45572.2020.9093349
Senushkin D, Romanov M, Belikov I et al (2021) Decoder modulation for indoor depth completion. In: IEEE/RSJ international conference on intelligent robots and systems, pp 2181–2188. https://doi.org/10.1109/IROS51168.2021.9636870
Deng Y, Deng X, Xu M (2023) A two-stage hybrid cnn-transformer network for rgb guided indoor depth completion. In: IEEE international conference on multimedia & expo (ICME), pp 1127–1132. https://doi.org/10.1109/ICME55011.2023.00197
Cheng X, Wang P, Yang R (2018) Depth estimation via affinity learned with convolutional spatial propagation network. In: European conference on computer vision (ECCV), Lecture Notes in Computer Science, vol 11220. Springer, pp 108–125. https://doi.org/10.1007/978-3-030-01270-0_7
Cheng X, Wang P, Guan C et al (2020) CSPN++: learning context and resource aware convolutional spatial propagation networks for depth completion. In: The AAAI conference on artificial intelligence. AAAI Press, pp 10615–10622. https://doi.org/10.1609/AAAI.V34I07.6635
Xu Z, Yin H, Yao J (2020) Deformable spatial propagation networks for depth completion. In: IEEE international conference on image processing (ICIP). IEEE, pp 913–917. https://doi.org/10.1109/ICIP40778.2020.9191138
Park J, Joo K, Hu Z et al (2020) Non-local spatial propagation network for depth completion. In: European conference on computer vision (ECCV), Lecture Notes in Computer Science, vol 12358. Springer, pp 120–136. https://doi.org/10.1007/978-3-030-58601-0_8
Zhao Y, Bai L, Zhang Z et al (2021) A surface geometry model for lidar depth completion. IEEE Robot Autom Lett 6(3):4457–4464. https://doi.org/10.1109/LRA.2021.3068885
Ren D, Yang M, Wu J et al (2023) Surface normal and gaussian weight constraints for indoor depth structure completion. Pattern Recognit 138:109362. https://doi.org/10.1016/j.patcog.2023.109362
Zhu Y, Dong W, Li L et al (2022) Robust depth completion with uncertainty-driven loss functions. In: The AAAI conference on artificial intelligence. AAAI Press, pp 3626–3634. https://doi.org/10.1609/AAAI.V36I3.20275
Chen R, Liu I, Yang E et al (2023) Activezero++: mixed domain learning stereo and confidence-based depth completion with zero annotation. IEEE Trans Pattern Anal Mach Intell 45(12):14098–14113. https://doi.org/10.1109/TPAMI.2023.3305399
Zhao T, Pan S, Zhang H et al (2021) Dilated u-block for lightweight indoor depth completion with sobel edge. IEEE Signal Process Lett 28:1615–1619. https://doi.org/10.1109/LSP.2021.3092280
Ramesh AN, Giovanneschi F, González-Huici MA (2023) Siunet: sparsity invariant u-net for edge-aware depth completion. In: IEEE winter conference on applications of computer vision (WACV). IEEE, pp 5807–5816. https://doi.org/10.1109/WACV56688.2023.00577
Jeon Y, Kim H, Seo SW (2021) ABCD: attentive bilateral convolutional network for robust depth completion. IEEE Robot Autom Lett 7(1):81–87. https://doi.org/10.1109/LRA.2021.3117254
Liu X, Shao X, Wang B et al (2022) Graphcspn: geometry-aware depth completion via dynamic gcns. In: European conference on computer vision (ECCV), Lecture Notes in Computer Science, vol 13693. Springer, pp 90–107. https://doi.org/10.1007/978-3-031-19827-4_6
Yu Z, Sheng Z, Zhou Z et al (2023) Aggregating feature point cloud for depth completion. In: International conference on computer vision (ICCV). IEEE, pp 8698–8709. https://doi.org/10.1109/ICCV51070.2023.00802
Zou Z, Chen K, Shi Z et al (2023) Object detection in 20 years: a survey. Proc IEEE 111(3):257–276. https://doi.org/10.1109/JPROC.2023.3238524
He K, Zhang X, Ren S et al (2016) Deep residual learning for image recognition. In: Conference on computer vision and pattern recognition (CVPR). IEEE Computer Society, pp 770–778. https://doi.org/10.1109/CVPR.2016.90
Shi W, Caballero J, Huszár F et al (2016) Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network. In: Conference on computer vision and pattern recognition (CVPR). IEEE Computer Society, pp 1874–1883. https://doi.org/10.1109/CVPR.2016.207
Silberman N, Hoiem D, Kohli P et al (2012) Indoor segmentation and support inference from rgbd images. In: European conference on computer vision (ECCV), vol 7576. Springer, pp 746–760. https://doi.org/10.1007/978-3-642-33715-4_54
Hu J, Shen L, Sun G (2018) Squeeze-and-excitation networks. In: Conference on computer vision and pattern recognition (CVPR). Computer Vision Foundation / IEEE Computer Society, pp 7132–7141. https://doi.org/10.1109/CVPR.2018.00745
Paszke A, Gross S, Massa F et al (2019) Pytorch: an imperative style, high-performance deep learning library. In: Advances in neural information processing systems, pp 8024–8035
Loshchilov I, Hutter F (2019) Decoupled weight decay regularization. In: International conference on learning representations (ICLR). OpenReview.net
Acknowledgements
This research is funded by the Science and Technology Commission of Shanghai Municipality (20511105102), The computation is performed in ECNU Multifunctional Platform for Innovation (001).
Funding
This research is funded by the Science and Technology Commission of Shanghai Municipality (20511105102).
Author information
Authors and Affiliations
Contributions
Zhichao Fu contributed to the design and implementation of the research, analyze the results, and wrote the main manuscript text. Xin Li prepared Figs. 1-2 and wrote the manuscript. Tianyu Huai and Weijie Li prepared the experiment results and wrote the manuscript. Daoguo Dong and Liang He devised the project and supervised the research. All authors reviewed the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors have no competing interests to declare that are relevant to the content of this article.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Fu, Z., Li, X., Huai, T. et al. Robust depth completion based on Semantic Aggregation. Appl Intell 54, 3825–3840 (2024). https://doi.org/10.1007/s10489-024-05366-5
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10489-024-05366-5