face recognition memory with nlp backend of...

26
Face Recognition Memory with NLP backend of Social Robot for HRI 8 th June 2018, Friday Nidhi Mishra Project Officer, IMI IMI Research Seminar 1

Upload: others

Post on 02-Jan-2020

4 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

Face Recognition Memory with NLP backend of Social Robot for HRI

8th June 2018, Friday

Nidhi Mishra

Project Officer, IMI

IMI Research Seminar

1

Page 2: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

Contents• Background of Robots

• Social Robots

• Motivation

• Proposed Model

• Algorithm

• Conclusion

• Future Work

• References

2

Page 3: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

3

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Industrial Service~27.5M 2

~1.7M

- https://assets.kpmg.com/content/dam/kpmg/pdf/2016/06/social-robots.pdf

Page 4: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

4

70%

17%

9%2%2%

Robots

Cleaning work Hobby Research Lawn-mowing Others

Hobby

Non-Social Social

Research

Non-Social Social

- https://assets.kpmg.com/content/dam/kpmg/pdf/2016/06/social-robots.pdf

Page 5: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

Robot roll-call

5

- https://assets.kpmg.com/content/dam/kpmg/pdf/2016/06/social-robots.pdf

Page 6: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

SocialHumanoid Robot

6

Page 7: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

Aging population

7

• The world is aging at a rapid rate and by 2030 there will be 34 nations where more than 20% of the population is over 65. This has broad implications for economic growth and immigration trends.

• With an older population that works less, support and dependency ratios get out of whack.

• Robots will perform many elder-care tasks within a decade.

- http://money.cnn.com/interactive/news/aging-countries/

Page 8: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

8

Page 9: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

Robot as Companion

Companion robots are robots that can make themselves useful, carry out various type of tasks to assist humans in a domestic environment. They should behave socially and interact with a socially acceptable manner with humans [Dautenhahn et al., 2005]

9

Page 10: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

Challenges

Several research have suggested characteristics that a social robot should visibly possess [Fong et al., 2003, Dautenhahn, 2007b, Heerink, 2010]:

• Express and perceive emotions

• Communicate with high level dialogue

• Learn and recognise models of other agents

• Establish and maintain social relationships

• Use natural cues of communication

• Exhibit distinctive personality and character

• Learn and develop social competencies

10

Page 11: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

Have a MemoryRecognize it’s CompanionUnderstand Human Language Naturally

11

Page 12: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

12

Proposed Solution

Page 13: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

Real Time face Recognition and Training

13

Page 14: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

OpenFace

• Detect faces with a pre-trained models from dlib or OpenCV.

• Transform the face for the neural network. This repository uses dlib's real-time pose estimation with OpenCV's affine transformation to try to make the eyes and bottom lip appear in the same location on each image.

• Use a deep neural network to represent (or embed) the face on a 128-dimensional unit hypersphere.

• Apply classification technique to the features to complete recognition task.

- https://cmusatyalab.github.io/openface/

14

Page 15: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

15

Page 16: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

16

Page 17: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

17

Page 18: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

18

Page 19: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

Extracting NLP features

• Part of speech (reads a sentence and assigns parts of speech to each word)

• Named entity recognition (labels sequences of words in a text which are the names of things, such as person and company names, or gene and protein names.)

• Find dependencies (representation of grammatical relations between words in a sentence)

• Coreference ( the task of finding all expressions that refer to the same entity in a text.)

-https://stanfordnlp.github.io/CoreNLP/index.html

19

Page 20: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

Sentence as input :-

1. Nidhi is working at NTU Singapore.

2. She works on Nadine robot.- http://nlp.stanford.edu:8080/corenlp/process

20

Page 21: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

WordNet

• WordNet is lexical database for English Language based conceptual look-up.

• Organizes lexical information in terms of word meanings rather than word form.

How wordNet can help us?One person can say a sentence in different ways using different words whose meaning are same.Example : I am going to downtown this weekend.

I am going to downtown this sunday.I am going to downtown this week.

WordNet can help me find similar words for one word. -http://www.nltk.org/howto/wordnet.html

21

Page 23: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

23

https://stanfordnlp.github.io/CoreNLP/index.htmlFabrizio Sebastiani. Machine Learning in Automated Text Categorization. ACM Computing Surveys, 34(1):1–47, Mars

Page 24: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

Conclusion

• For robots, it is crucial that they possess human-like social interaction skills. It is vital for a robot to know who it is talking with and remember facts about the human companion.

• Aims to explore the possibility of improving human-robot interaction(HRI) by exploiting natural language resources and using natural language processing (NLP) methods.

24

Page 25: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

Future Plans

• An illustration through a real case of this model.

• Compare the model using standard database and a new database for HRI applications.

25

Page 26: Face Recognition Memory with NLP backend of …imi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/...Face Recognition Memory with NLP backend of Social Robot for HRI 8th June

Thank you!

Q & A

26