Robust Neural Routing Through Space Partitions for Camera Relocalization in Dynamic Indoor Environments (2020-12-08T00:00:00.000000Z)

TL;DR

This paper proposes a novel outlier-aware neural tree which bridges the two worlds, deep learning and decision tree approaches and outperforms the state-of-the-art approaches by around 30% on camera pose accuracy, while running comparably fast for evaluation.

Abstract

Localizing the camera in a known indoor environment is a key building block for scene mapping, robot navigation, AR, etc. Recent advances estimate the camera pose via optimization over the 2D/3D-3D correspondences established between the coordinates in 2D/3D camera space and 3D world space. Such a mapping is estimated with either a convolution neural network or a decision tree using only the static input image sequence, which makes these approaches vulnerable to dynamic indoor environments that are quite common yet challenging in the real world. To address the aforementioned issues, in this paper, we propose a novel outlier-aware neural tree which bridges the two worlds, deep learning and decision tree approaches. It builds on three important blocks: (a) a hierarchical space partition over the indoor scene to construct the decision tree; (b) a neural routing function, implemented as a deep classification network, employed for better 3D scene understanding; and (c) an outlier rejection module used to filter out dynamic points during the hierarchical routing process. Our proposed algorithm is evaluated on the RIO-10 benchmark developed for camera relocalization in dynamic indoor environments. It achieves robust neural routing through space partitions and outperforms the state-of-the-art approaches by around 30% on camera pose accuracy, while running comparably fast for evaluation.

Authors

L. Guibas

28 papers

T. Funkhouser

13 papers

He Wang

4 papers

TL;DR

Abstract

Authors

References60 items

Beyond Controlled Environments: 3D Camera Re-Localization in Changing Indoor Scenes

InLoc: Indoor Visual Localization with Dense Matching and View Synthesis

SANet: Scene Agnostic Network for Camera Localization

Hierarchical Scene Coordinate Classification and Regression for Visual Localization

AtLoc: Attention Guided Camera Localization

Expert Sample Consensus Applied to Camera Re-Localization

A General Decoupled Learning Framework for Parameterized Image Operators

Neural-Guided RANSAC: Learning Where to Sample Model Hypotheses

Understanding the Limitations of CNN-Based Absolute Camera Pose Regression

From Coarse to Fine: Robust Hierarchical Localization at Large Scale

Real-Time RGB-D Camera Pose Estimation in Novel Scenes Using a Relocalisation Cascade

Leveraging Deep Visual Descriptors for Hierarchical Efficient Localization

PPF-FoldNet: Unsupervised Learning of Rotation Invariant 3D Local Descriptors

Decouple Learning for Parameterized Image Operators

Full-Frame Scene Coordinate Regression for Image-Based Localization

24/7 Place Recognition by View Synthesis

Deep Regression Forests for Age Estimation

Geometry-Aware Learning of Maps for Camera Localization

Learning Less is More - 6D Camera Localization via 3D Surface Regression

Exploiting Points and Lines in Regression Forests for RGB-D Camera Relocalization

Backtracking regression forests for accurate camera relocalization

Efficient & Effective Prioritized Matching for Large-Scale Image-Based Localization

PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space

Self-Supervised Visual Descriptor Learning for Dense Correspondence

Geometric Loss Functions for Camera Pose Regression with Deep Learning

Label Distribution Learning Forests

On-the-Fly Adaptation of Regression Forests for Online Camera Relocalisation

PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation

A Point Set Generation Network for 3D Object Reconstruction from a Single Image

Image-Based Localization Using LSTMs for Structured Feature Correlation

DSAC — Differentiable RANSAC for Camera Localization

Random forests versus Neural Networks — What's best for camera localization?

Monocular Depth Estimation Using Neural Regression Forest

Large-Scale Location Recognition and the Geometric Burstiness Problem

Uncertainty-Driven 6D Pose Estimation of Objects and Scenes from a Single RGB Image

Learning to Navigate the Energy Landscape

Deep Neural Decision Forests

NetVLAD: CNN Architecture for Weakly Supervised Place Recognition

Exploiting uncertainty in regression forests for accurate camera relocalization

PoseNet: A Convolutional Network for Real-Time 6-DOF Camera Relocalization

Real-Time RGB-D Camera Relocalization via Randomized Ferns for Keyframe Encoding

Adam: A Method for Stochastic Optimization

Multi-output Learning for Camera Relocalization

Scene Coordinate Regression Forests for Camera Relocalization in RGB-D Images

Real-time image-based 6-DOF localization in large-scale environments

Real-time loop detection with bags of binary words

Fast image-based localization using direct 2D-to-3D matching

Aggregating local descriptors into a compact image representation

Locally Optimized RANSAC

Mean Shift: A Robust Approach Toward Feature Space Analysis

Top-Down Induction of Clustering Trees

Surface reconstruction from unorganized points

Method for registration of 3-D shapes

A Method for Registration of 3-D Shapes

A solution for the best rotation to relate two sets of vectors

LIII. On lines and planes of closest fit to systems of points in space

D2-net: A trainable cnn for joint detection and description of local features

Clustering with Decision Trees: Divisive and Agglomerative Approach

6D Relocalisation for RGBD Cameras Using Synthetic View Regression

on lines and planes of closest fit to systems of points in space. The London, Edinburgh, and Dublin Philosophical Magazine and

Field of Study

Journal Information

Name

Page

Venue Information

Name

Type

URL

Alternate Names