High Quality Monocular Depth Estimation via Transfer Learning

Abstract

Accurate depth estimation from images is a fundamental task in many applications including scene understanding and reconstruction. Existing solutions for depth estimation often produce blurry approximations of low resolution. This paper presents a convolutional neural network for computing a high-resolution depth map given a single RGB image with the help of transfer learning. Following a standard encoder-decoder architecture, we leverage features extracted using high performing pre-trained networks when initializing our encoder along with augmentation and training strategies that lead to more accurate results. We show how, even for a very simple decoder, our method is able to achieve detailed high-resolution depth maps. Our network, with fewer parameters and training iterations, outperforms state-of-the-art on two datasets and also produces qualitatively better results that capture object boundaries more faithfully. Code and corresponding pre-trained weights are made publicly available.

arXiv comment12 pages

References in corpus (6)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Depth Map Prediction from a Single Image using a Multi-Scale Deep Network
• Benchmark Analysis of Representative Deep Neural Network Architectures
• DeMoN: Depth and Motion Network for Learning Monocular Stereo
• AutoAugment: Learning Augmentation Policies from Data
• Detail Preserving Depth Estimation from a Single Image Using Attention Guided Networks

Cited by in corpus (47)

• AdaBins: Depth Estimation using Adaptive Bins
• DeepV2D: Video to Depth with Differentiable Structure from Motion
• On Deep Learning Techniques to Boost Monocular Depth Estimation for Autonomous Navigation
• DenseLiDAR: A Real-Time Pseudo Dense Depth Guided Depth Completion Network
• DIODE: A Dense Indoor and Outdoor DEpth Dataset
• RefinedMPL: Refined Monocular PseudoLiDAR for 3D Object Detection in Autonomous Driving
• Absolute distance prediction based on deep learning object detection and monocular depth estimation models
• DiverseDepth: Affine-invariant Depth Prediction Using Diverse Data
• ClearGrasp: 3D Shape Estimation of Transparent Objects for Manipulation
• Regression Prior Networks
• Fast and Accurate Single-Image Depth Estimation on Mobile Devices, Mobile AI 2021 Challenge: Report
• M^3VSNet: Unsupervised Multi-metric Multi-view Stereo Network
• Dense Uncertainty Estimation
• Conf-Net: Toward High-Confidence Dense 3D Point-Cloud with Error-Map Prediction
• Rethinking Pseudo-LiDAR Representation
• Bidirectional Attention Network for Monocular Depth Estimation
• Task-Aware Monocular Depth Estimation for 3D Object Detection
• Deep Learning based Monocular Depth Prediction: Datasets, Methods and Applications
• DepthNet Nano: A Highly Compact Self-Normalizing Neural Network for Monocular Depth Estimation
• Deep Shape-from-Template: Wide-Baseline, Dense and Fast Registration and Deformable Reconstruction from a Single Image
• Towards Real-Time Monocular Depth Estimation for Robotics: A Survey
• Resonant Scanning Design and Control for Fast Spatial Sampling
• Pseudo Supervised Monocular Depth Estimation with Teacher-Student Network
• Discovering Supply Chain Links with Augmented Intelligence
• Toward Hierarchical Self-Supervised Monocular Absolute Depth Estimation for Autonomous Driving Applications
• MSFNet:Multi-scale features network for monocular depth estimation
• Virtual Normal: Enforcing Geometric Constraints for Accurate and Robust Depth Prediction
• Progressive Depth Learning for Single Image Dehazing
• Towards Panoptic 3D Parsing for Single Image in the Wild
• Depth-only Object Tracking
• Improving Monocular Depth Estimation by Leveraging Structural Awareness and Complementary Datasets
• Low-viewpoint forest depth dataset for sparse rover swarms
• A Weakly-Supervised Depth Estimation Network Using Attention Mechanism
• Echo-Reconstruction: Audio-Augmented 3D Scene Reconstruction
• Multimodal Data Fusion for Power-On-and-Go Robotic Systems in Retail
• Mirror3D: Depth Refinement for Mirror Surfaces
• Focusing on Shadows for Predicting Heightmaps from Single Remotely Sensed RGB Images with Deep Learning
• Dark Reciprocal-Rank: Boosting Graph-Convolutional Self-Localization Network via Teacher-to-student Knowledge Transfer
• Depth estimation on embedded computers for robot swarms in forest
• Parallel Blockwise Knowledge Distillation for Deep Neural Network Compression
• Improved Point Transformation Methods For Self-Supervised Depth Prediction
• Object Detection on Single Monocular Images through Canonical Correlation Analysis
• Topological Regularization for Dense Prediction
• A Novel 3D-UNet Deep Learning Framework Based on High-Dimensional Bilateral Grid for Edge Consistent Single Image Depth Estimation
• Single Image Depth Prediction with Wavelet Decomposition
• NeRD: Neural 3D Reflection Symmetry Detector
• Virtual Multi-Modality Self-Supervised Foreground Matting for Human-Object Interaction