classification of five chinese tea categories with

36

Upload: others

Post on 18-Jan-2022

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Classification of five Chinese tea categories with
Page 2: Classification of five Chinese tea categories with

2

Page 3: Classification of five Chinese tea categories with

3

Page 4: Classification of five Chinese tea categories with

4

Page 5: Classification of five Chinese tea categories with

5

2D 3D

• Immersive experience

• More Application

Page 6: Classification of five Chinese tea categories with

6

How to inpainting

Page 7: Classification of five Chinese tea categories with

7

Page 8: Classification of five Chinese tea categories with

8

• Multi-plane representation

• Learning-based

• Facebook => Layered depth image (LDI)

Page 9: Classification of five Chinese tea categories with

9

d = 1

d = 2

• Color and depth value

• Holds any number of pixels

New Rule:

• No neighbors across depth

discontinuities

Page 10: Classification of five Chinese tea categories with

10

Page 11: Classification of five Chinese tea categories with

11

RGB – D data

Image inpaintingalgorithm

3D photography

Page 12: Classification of five Chinese tea categories with

12

RGB – D data

Image inpaintingalgorithm

3D photography

Page 13: Classification of five Chinese tea categories with

13

9 10 20

5 3 30

40 50 100

Depth map

0.09 0.1 0.2

0.05 0.03 0.3

0.4 0.5 1

Normalizeto 0 ~ 1

Bilateral median filter

window 7 x 7

spatial 4.0

intensity 0.5

Page 14: Classification of five Chinese tea categories with

14

Raw Filtered

Page 15: Classification of five Chinese tea categories with

15

ThresholdCheck

Connectivity

Page 16: Classification of five Chinese tea categories with

16

RGB – D data

Image inpaintingalgorithm

3D photography

Page 17: Classification of five Chinese tea categories with

17

Page 18: Classification of five Chinese tea categories with

18

Page 19: Classification of five Chinese tea categories with

19

RGB – D data

Image inpaintingalgorithm

3D photography

Page 20: Classification of five Chinese tea categories with

20

Page 21: Classification of five Chinese tea categories with

21

• Edge inpainting network:GAN

• Depth inpainting network:U-net

• Color inpainting network:U-net

Color inpainting loss

Depth inpainting loss

Page 22: Classification of five Chinese tea categories with

22

RGB – D data

Image inpaintingalgorithm

3D photography

Page 23: Classification of five Chinese tea categories with

23

Page 24: Classification of five Chinese tea categories with

24

RGB – D data

Image inpaintingalgorithm

3D photography

Page 25: Classification of five Chinese tea categories with

25

RGB – D data LDI image Mesh representation

Page 26: Classification of five Chinese tea categories with

26

Page 27: Classification of five Chinese tea categories with

27

Page 28: Classification of five Chinese tea categories with

28

Page 29: Classification of five Chinese tea categories with

29

Facebook Proposed

Page 30: Classification of five Chinese tea categories with

30

Page 31: Classification of five Chinese tea categories with

31

Page 32: Classification of five Chinese tea categories with

32

Page 33: Classification of five Chinese tea categories with

33

Page 34: Classification of five Chinese tea categories with

34

• Proposed an algorithm to create 3D photography from RGB-D image

• Created layered depth image representation through context-aware

color and depth inpainting

• Applied edge inpainting network to inpaint occlusion edges

• Produced fewer artifacts when compared with other techniques

Page 35: Classification of five Chinese tea categories with

35

1. 3D Photography using Context-aware Layered Depth Inpainting. Meng-Li Shih,

Shih-Yang Su, Johannes Kopf, Jia-Bin Huang. CVPR, 2020.

2. EdgeConnect: Generative Image Inpainting with Adversarial Edge Learning.

Kamyar Nazeri, Eric Ng, Tony Joseph, Faisal Z. Qureshi, Mehran Ebrahimi. ICCV,

2019.

3. The unreasonable effectiveness of deep features as a perceptual metric. In

CVPR, 2018

Page 36: Classification of five Chinese tea categories with

36