super resolution on 3d point clouds using deep learning€¦ · master in computer vision ......

22
Super Resolution on 3D Point Clouds using Deep Learning Belén Luque López

Upload: others

Post on 21-May-2020

5 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Super Resolution on 3D Point

Clouds

using Deep Learning

Belén Luque López

Page 2: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

About me

Bachelor's Degree in

Audiovisual Systems

Engineering

Master in Computer

Vision

Master’s Thesis

Javier Ruiz Béatrice

Pesquet

Super Resolution on 3D

Point Clouds using Deep

Learning

May - September

2017

2

Page 3: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Motivation

Telepresence project

Internship at the Image Processing Group

of the UPC (September 2016-April 2017)

Capture room

Visualization room

3

Page 4: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Motivation

Low resolution of point clouds!

Possible solutions:

● Use meshes

● Increase the number

of points 4

Telepresence project

Page 5: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Working with 3D

2.5D (RGB-

D)

Multi-view

projection

Voxelizatio

n

Point cloud

Unorganized list of XYZ

coordinates 5

Page 6: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Super resolution

Super resolution in 2D Super resolution in 2.5D

Super resolution on depth map, then construct

point cloud

Upsample image without losing

spatial detail 6

Page 7: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Super resolution

Super resolution in 3D Super resolution in 2D

7

Create more points, fill the holes

Page 8: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Super resolution

Super resolution in 3D Super resolution in 2D

The position of the new points is

already given

Multiple options, not that

easy! 8

Page 9: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

C. R. Qi, H. Su, K. Mo, and L. J. Guibas, “PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation,” CoRR, vol. abs/1612.00593, 2016.

Our first approach

Pointne

t

Novel deep net architecture that directly consumes point clouds

(unordered point sets)

9

Page 10: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Our first approach

The input is an unordered list of XYZ

coordinates!

10

Pointne

t

Page 11: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Our first approach

11

Pointne

t

mini network to learn affine transformation matrix

make the point cloud invariant to rotation/translation

Page 12: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Our first approach

12

Pointne

t

Page 13: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Our first approach

13

Pointne

t X

Y

Z ...

64

XYZ

XYZ

XYZ

n n

Page 14: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Our first approach

14

Pointne

t

Page 15: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Our first approach

15

Pointne

t

10 5 1 3

1 8 5 9

6 2 3 5

9 1 6 3

4 9 3 5

10 9 6 9

Page 16: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Our first approach

16

Pointne

t

Page 17: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Our first approach

Over 6,000 m2 of indoor spaces, almost 700.000.000 points + meshes

Dataset: 2D-3D-Semantics

(http://buildingparser.stanford.edu/dataset.html)

I. Armeni, A. Sax, A. R. Zamir, and S. Savarese,

“Joint 2D-3D-Semantic Data for Indoor Scene

Understanding” ArXiv e-prints, 2017.

17

Page 18: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Our first approach

Input (N

points)

Output (N

points)

Original point cloud (2N

points)

Pointnet

Original point

cloud

50% sampling

Input

data

Ground

truth

Residual training to obtain the double of points

50%

sampling

18

Page 19: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Our first approach

Pointnet

nx

3

Loss:

1. For each output point, find the closest point in the ground truth (nearest neighbor

search with kd-tree)

2. Minimize RMS distance

Output: new XYZ

coordinates

19

Page 20: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Tools

● Data preparation (point cloud downsampling):

PCL library

● Pointnet model: Tensorflow (or third-party

implementation in pyTorch)

○ + Integration with python-pcl for point cloud

processing (nearest neighbor search)

● Data ingestion: h5py library (dataset stored in

HDF5 format)

● Data augmentation (rotation/jittering): numpy

library

● Visualization of results: meshlab software

numpy + scipy 2D

projection

20

Page 21: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Some thoughts

Can we combine the input points (with convolutions, like in 2D images) ?

The input is an unorganized list of 3D points, neighborhoods are not

defined…

→ Semantic segmentation as a previous step?

Multiple solutions exist for the creation of new points. How do we train the

network?

- Minimize distance to closest point in the ground truth

- Minimize distance to the surface (closest face in a mesh)

Do we want the network to create more

points in the less populated areas? How? 21

Page 22: Super Resolution on 3D Point Clouds using Deep Learning€¦ · Master in Computer Vision ... Javier Ruiz Béatrice Pesquet Super Resolution on 3D Point Clouds using Deep Learning

Thank you!

Belén Luque López