综述自动驾驶中的计算机视觉Computer Vision for Autonomous Vehicles: Problems, Datasets and State-of-the-Art（下）

6. Semantic Segmentation

Formulation

Structured CNNs

Conditional Random Fields

Discussion

6.1 Semantic Instance Segmentation

Proposal-based Instance Segmentation

Proposal-free Instance Segmentation

Discussion

6.2 Label Propagation

6.3 Semantic Segmentation with Multiple Frames

6.4 Semantic Segmentation of 3D dATA

Online Methods

3D CNN

6.5 Semantic Segmentation of Street Side Views

6.6 Semantic Segmentation of Aerial Images

Aerial Image Parsing using Maps

Fine-grained Image Parsing with Aerial-to-ground Reasoning

6.6.1 ISPRS Segmentation Challenge

6.7 Road Segmentation

CNN-based Methods

6.7.1 Free Space Estimation

Long Range Obstacle Detection

7. Reconstruction

7.1 Stereo

Taxonomies

Matching Cost Functions

SGM

Variable Baseline/Resolution

Planarity

Variational Approaches

State-of-the-art

Superpixels

Deep Learning

Discussion

7.2 Multi-view 3D Reconstruction

Taxonomies

Representations:Depth Map

Representations:Point-cloud

Representations:Volumetric

Representations:Mesh or Surface

Urban Reconstruction

Input Data

Stereo Sequences

Digital Surface Models(DSM)

Air- and Street-level

Stereo Satellite

7.3 Reconstruction and Recognition

Planarity and Primitives

Volumetric

Volumetric:Large-scale

Shape Priors

Data-Driven

8. Motion & Pose Estimation

8.1 2D Motion Estimation - Optical Flow

Variational Formulation

Sparse Matches

High Speed Flow

State-of-the-art

Epipolar Flow

Semantic Segmentation

Confidences

Deep Learning

Discussion

8.2 3D Motion Estimation - Scene Flow

Variational Approaches

Piecewise Rigidity

Piecewise Rigidity at the Object Level

State-of-the-art

Discussion

8.3 Ego-Motion Estimation

Formulation

Drift

2D-to-2D Matching

3D-to-2D Matching

3D-to-3D Matching

8.3.1 State-of-the-art

Monocular Visual Odometry

Stereo Visual Odometry

LiDAR-base Odometry

Discussion

8.4 Simultaneous Localization and Mapping(SLAM)

Formulation

Environmental Changes

8.4.1 Loop Closure Detection

LiDAR-based

8.4.2 Visual SLAM

8.4.3 Mapping

Metric Maps

Semantic Maps

8.5 Localization

Survey

Monte Carlo Methods

Metric, Topological, Topometric

Scale and Accuracy

Structured-based Localization

Structured-based Localization using Deep Learning

Cross-view Localization

Cross-view Localization:Buildings

Cross-view Localization:Reconstructions

Semantic Alignment from LiDAR

9. Tracking

Challenges

Formulation

On MOT16

On KITTI

未完~