Monocular Depth Estimation Primed by Salient Point Detection and Normalized Hessian Loss
Not on the map. Why?
Abstract (arXiv)
Deep neural networks have recently thrived on single image depth estimation. That being said, current developments on this topic highlight an apparent compromise between accuracy and network size. This work proposes an accurate and lightweight framework for monocular depth estimation based on a self-attention mechanism stemming from salient point detection. Specifically, we utilize a sparse set of keypoints to train a FuSaNet model that consists of two major components: Fusion-Net and Saliency-Net. In addition, we introduce a normalized Hessian loss term invariant to scaling and shear along the depth direction, which is shown to substantially improve the accuracy. The proposed method achieves state-of-the-art results on NYU-Depth-v2 and KITTI while using 3.1-38.4 times smaller model in terms of the number of parameters than baseline approaches. Experiments on the SUN-RGBD further demonstrate the generalizability of the proposed method.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 51 of its 71 references that are in the map. First posted on arXiv 25 Aug 2021.
Builds on
51 of 71 references.
References · 51
- 2014Adam: A Method for Stochastic Optimizationcited by 36,960
- 2017Squeeze-and-Excitation Networkscited by 4,763
- 2017Non-local Neural Networkscited by 2,290
- 2021Vision Transformers for Dense Predictioncited by 1,789
- 2015Show, Attend and Tell: Neural Image Caption Generation with Visual Attentioncited by 1,421
- 2014Depth Map Prediction from a Single Image using a Multi-Scale Deep Networkcited by 1,264
- 2018MnasNet: Platform-Aware Neural Architecture Search for Mobilecited by 937
- 2018Deep Ordinal Regression Network for Monocular Depth Estimationcited by 721
- 2014Predicting Depth, Surface Normals and Semantic Labels with a Common Multi-Scale Convolutional Architecturecited by 483
- 2016Deeper Depth Prediction with Fully Convolutional Residual Networkscited by 402
- 2017Residual Attention Network for Image Classificationcited by 385
- 2019From Big to Small: Multi-Scale Local Planar Guidance for Monocular Depth Estimationcited by 379
38 more
- 2019Attention Augmented Convolutional Networkscited by 363
- 2017Sparsity Invariant CNNscited by 325
- 2019Enforcing geometric constraints of virtual normal for depth predictioncited by 233
- 2020Non-Local Spatial Propagation Network for Depth Completioncited by 194
- 2017Sparse-to-Dense: Depth Prediction from Sparse Depth Samples and a Single Imagecited by 159
- 2021PENet: Towards Precise and Efficient Image Guided Depth Completioncited by 146
- 2018Revisiting Single Image Depth Estimation: Toward Higher Resolution Maps with Accurate Object Boundariescited by 144
- 2019Pattern-Affinitive Propagation across Depth, Surface Normal and Semantic Segmentationcited by 141
- 2018Learning Depth with Convolutional Spatial Propagation Networkcited by 135
- 2018DeepLiDAR: Deep Surface Normal Guided Depth Prediction for Outdoor Scene from Sparse LiDAR Data and Single Color Imagecited by 129
- 2018Deep Depth Completion of a Single RGB-D Imagecited by 110
- 2018Depth Estimation via Affinity Learned with Convolutional Spatial Propagation Networkcited by 109
- 2019Depth Completion from Sparse LiDAR Data with Depth-Normal Constraintscited by 108
- 2018PlaneRCNN: 3D Plane Detection and Reconstruction from a Single Imagecited by 105
- 2019FastDepth: Fast Monocular Depth Estimation on Embedded Systemscited by 93
- 2020Guiding Monocular Depth Estimation Using Depth-Attention Volumecited by 82
- 2018Structured Attention Guided Convolutional Neural Fields for Monocular Depth Estimationcited by 78
- 2018PlaneNet: Piece-wise Planar Reconstruction from a Single RGB Imagecited by 77
- 2018Sparse and Dense Data with CNNs: Depth Completion and Semantic Segmentationcited by 76
- 2019CAM-Convs: Camera-Aware Multi-Scale Convolutions for Single-View Depthcited by 75
- 2020Learning Joint 2D-3D Representations for Depth Completioncited by 71
- 2019SharpNet: Fast and Accurate Recovery of Occluding Contours in Monocular Depth Estimationcited by 59
- 2020Uncertainty-Aware CNNs for Depth Completion: Uncertainty from Beginning to Endcited by 55
- 2018Confidence Propagation through CNNs for Guided Sparse Depth Regressioncited by 52
- 2020Forget About the LiDAR: Self-Supervised Depth Estimators with MED Probability Volumescited by 50
- 2019Structure-Aware Residual Pyramid Network for Monocular Depth Estimationcited by 49
- 2019Depth Coefficients for Depth Completioncited by 41
- 2018Propagating Confidences through CNNs for Sparse Data Regressioncited by 36
- 2018Deep attention-based classification network for robust depth predictioncited by 36
- 2016Physically-Based Rendering for Indoor Scene Understanding Using Convolutional Neural Networkscited by 35
- 2018Detail Preserving Depth Estimation from a Single Image Using Attention Guided Networkscited by 31
- 2018Towards real-time unsupervised monocular depth estimation on CPUcited by 26
- 2017The Devil is in the Decoder: Classification, Regression and GANscited by 22
- 2021Fast and Accurate Single-Image Depth Estimation on Mobile Devices, Mobile AI 2021 Challenge: Reportcited by 19
- 2019Deep Robust Single Image Depth Estimation Neural Network Using Scene Understandingcited by 15
- 2020Boosting Monocular Depth Estimation with Lightweight 3D Point Fusioncited by 14
- 2018Pixel-wise Attentional Gating for Parsimonious Pixel Labelingcited by 12
- 2021Probabilistic Graph Attention Network with Conditional Kernels for Pixel-Wise Predictioncited by 8
50 most cited of 51 All 51 (JSON)
Built on by
Citing papers per year: 2022: 1.
1 paper.