다중 곡면 검출 및 추적을 이용한 증강현실 책

Similar documents
(JBE Vol. 24, No. 4, July 2019) (Special Paper) 24 4, (JBE Vol. 24, No. 4, July 2019) ISSN

<4D F736F F D20B1E2C8B9BDC3B8AEC1EE2DC0E5C7F5>

(JBE Vol. 24, No. 2, March 2019) (Special Paper) 24 2, (JBE Vol. 24, No. 2, March 2019) ISSN

(JBE Vol. 23, No. 2, March 2018) (Special Paper) 23 2, (JBE Vol. 23, No. 2, March 2018) ISSN

(JBE Vol. 22, No. 2, March 2017) (Special Paper) 22 2, (JBE Vol. 22, No. 2, March 2017) ISSN

(JBE Vol. 23, No. 2, March 2018) (Special Paper) 23 2, (JBE Vol. 23, No. 2, March 2018) ISSN

07.045~051(D04_신상욱).fm

09오충원(613~623)

Æí¶÷4-¼Ö·ç¼Çc03ÖÁ¾š

02( ) SAV12-19.hwp

High Resolution Disparity Map Generation Using TOF Depth Camera In this paper, we propose a high-resolution disparity map generation method using a lo

À±½Â¿í Ãâ·Â

09권오설_ok.hwp

°í¼®ÁÖ Ãâ·Â

ȲÀμº Ãâ·Â

Ch 1 머신러닝 개요.pptx

<4D F736F F D20C3D6BDC C0CCBDB4202D20BAB9BBE7BABB>

Reinforcement Learning & AlphaGo

1. 서 론

보고싶었던 Deep Learning과 OpenCV를이용한이미지처리과정에대해공부를해볼수있으며더나아가 Deep Learning기술을이용하여논문을작성하는데많은도움을받을수있으며아직배우는단계에있는저에게는기존의연구를따라해보는것만으로도큰발전이있다고생각했습니다. 그래서이번 DSP스마

融合先验信息到三维重建 组会报 告[2]

(JBE Vol. 23, No. 5, September 2018) (Special Paper) 23 5, (JBE Vol. 23, No. 5, September 2018) ISSN

4 : (Hyo-Jin Cho et al.: Audio High-Band Coding based on Autoencoder with Side Information) (Special Paper) 24 3, (JBE Vol. 24, No. 3, May 2019

±¹Á¦ÆòÈŁ4±Ç1È£-ÃÖÁ¾

PowerPoint Presentation

PowerPoint 프레젠테이션

2 : (EunJu Lee et al.: Speed-limit Sign Recognition Using Convolutional Neural Network Based on Random Forest). (Advanced Driver Assistant System, ADA

Microsoft PowerPoint - 실습소개와 AI_ML_DL_배포용.pptx

6 : (Gicheol Kim et al.: Object Tracking Method using Deep Learing and Kalman Filter) (Regular Paper) 24 3, (JBE Vol. 24, No. 3, May 2019) http

The Third NTCIR Workshop, Sep

3 : 3D (Seunggi Kim et. al.: 3D Depth Estimation by a Single Camera) (Regular Paper) 24 2, (JBE Vol. 24, No. 2, March 2019)

26 이경승(394~400).hwp

2 : (Seungsoo Lee et al.: Generating a Reflectance Image from a Low-Light Image Using Convolutional Neural Network) (Regular Paper) 24 4, (JBE

09È«¼®¿µ 5~152s

Ⅱ. Embedded GPU 모바일 프로세서의 발전방향은 저전력 고성능 컴퓨팅이다. 이 러한 목표를 달성하기 위해서 모바일 프로세서 기술은 멀티코 어 형태로 발전해 가고 있다. 예를 들어 NVIDIA의 최신 응용프 로세서인 Tegra3의 경우 쿼드코어 ARM Corte

DBPIA-NURIMEDIA

2 : (JEM) QTBT (Yong-Uk Yoon et al.: A Fast Decision Method of Quadtree plus Binary Tree (QTBT) Depth in JEM) (Special Paper) 22 5, (JBE Vol. 2

2 247, Dec.07, 2007

DBPIA-NURIMEDIA

SuaKITBrochure_v2.2_KO

¼º¿øÁø Ãâ·Â-1

Delving Deeper into Convolutional Networks for Learning Video Representations - Nicolas Ballas, Li Yao, Chris Pal, Aaron Courville arXiv:

Artificial Intelligence: Assignment 6 Seung-Hoon Na December 15, Sarsa와 Q-learning Windy Gridworld Windy Gridworld의 원문은 다음 Sutton 교재의 연습문제

2 : (Juhyeok Mun et al.: Visual Object Tracking by Using Multiple Random Walkers) (Special Paper) 21 6, (JBE Vol. 21, No. 6, November 2016) ht

<91E6308FCD5F96DA8E9F2E706466>

R을 이용한 텍스트 감정분석

DBPIA-NURIMEDIA

Microsoft Word - retail_ doc

<BACFC7D1B3F3BEF7B5BFC7E22D3133B1C733C8A BFEB2E687770>


Visual recognition in the real world SKT services

Electronics and Telecommunications Trends 인공지능을이용한 3D 콘텐츠기술동향및향후전망 Recent Trends and Prospects of 3D Content Using Artificial Intelligence Technology

지능정보연구제 16 권제 1 호 2010 년 3 월 (pp.71~92),.,.,., Support Vector Machines,,., KOSPI200.,. * 지능정보연구제 16 권제 1 호 2010 년 3 월


(JBE Vol. 20, No. 5, September 2015) (Special Paper) 20 5, (JBE Vol. 20, No. 5, September 2015) ISS

04 김영규.hwp

2 : (Minsong Ki et al.: Lower Tail Light Learning-based Forward Vehicle Detection System Irrelevant to the Vehicle Types) (Regular) 21 4, (JBE

(JBE Vol. 23, No. 4, July 2018) (Special Paper) 23 4, (JBE Vol. 23, No. 4, July 2018) ISSN

김기남_ATDC2016_160620_[키노트].key

0125_ 워크샵 발표자료_완성.key

(JBE Vol. 24, No. 4, July 2019) (Special Paper) 24 4, (JBE Vol. 24, No. 4, July 2019) ISSN

170

006- 5¿ùc03ÖÁ¾T300çÃâ

THE JOURNAL OF KOREAN INSTITUTE OF ELECTROMAGNETIC ENGINEERING AND SCIENCE Jul.; 29(7),

歯3일_.PDF

박선영무선충전-내지

성능 감성 감성요구곡선 평균사용자가만족하는수준 성능요구곡선 성능보다감성가치에대한니즈가증대 시간 - 1 -

THE JOURNAL OF KOREAN INSTITUTE OF ELECTROMAGNETIC ENGINEERING AND SCIENCE Jul.; 27(7),

08/11-12<È£ä263»Áö

State of Play - Video Insights Report_Korean_v2.key

ch3.hwp

Gray level 변환 및 Arithmetic 연산을 사용한 영상 개선

1. 서 론

<465441C8B0BFEBC1F6BFF8BCBEC5CDC8ABBAB8C6D4C7C3B8B E696E6464>

3 : OpenCL Embedded GPU (Seung Heon Kang et al. : Parallelization of Feature Detection and Panorama Image Generation using OpenCL and Embedded GPU). e

<353420B1C7B9CCB6F52DC1F5B0ADC7F6BDC7C0BB20C0CCBFEBC7D120BEC6B5BFB1B3C0B0C7C1B7CEB1D7B7A52E687770>

À¯Çõ Ãâ·Â

정보기술응용학회 발표

±è¼ºÃ¶ Ãâ·Â-1

2 차원단위블록정렬을이용한 내용기반이미지매칭 장철진 O 조환규부산대학교컴퓨터공학과 {jin, Content-based image matching based on 2D alignment of unit block tessellation C

THE JOURNAL OF KOREAN INSTITUTE OF ELECTROMAGNETIC ENGINEERING AND SCIENCE Mar.; 29(3),

08/09-10;È£ä263»Áö


歯4차학술대회원고(황수경이상호).PDF

,. 3D 2D 3D. 3D. 3D.. 3D 90. Ross. Ross [1]. T. Okino MTD(modified time difference) [2], Y. Matsumoto (motion parallax) [3]. [4], [5,6,7,8] D/3

서강대학교영상처리연구실 서봉준 양나은 주혜진

PowerPoint Presentation

(JBE Vol. 7, No. 4, July 0)., [].,,. [4,5,6] [7,8,9]., (bilateral filter, BF) [4,5]. BF., BF,. (joint bilateral filter, JBF) [7,8]. JBF,., BF., JBF,.

(JBE Vol. 24, No. 1, January 2019) (Special Paper) 24 1, (JBE Vol. 24, No. 1, January 2019) ISSN 2287-

<4D F736F F D20C3D6BDC C0CCBDB4202D20BAB9BBE7BABB>

Journal of Educational Innovation Research 2019, Vol. 29, No. 1, pp DOI: (LiD) - - * Way to

1. GigE Camera Interface를 위한 최소 PC 사양 CPU : Intel Core 2 Duo, 2.4GHz이상 RAM : 2GB 이상 LANcard : Intel PRO/1000xT 이상 VGA : PCI x 16, VRAM DDR2 RAM 256MB

2011´ëÇпø2µµ 24p_0628

45-51 ¹Ú¼ø¸¸

PowerPoint 프레젠테이션

19_9_767.hwp

<35335FBCDBC7D1C1A42DB8E2B8AEBDBAC5CDC0C720C0FCB1E2C0FB20C6AFBCBA20BAD0BCAE2E687770>

KCC2011 우수발표논문 휴먼오피니언자동분류시스템구현을위한비결정오피니언형용사구문에대한연구 1) Study on Domain-dependent Keywords Co-occurring with the Adjectives of Non-deterministic Opinion

THE JOURNAL OF KOREAN INSTITUTE OF ELECTROMAGNETIC ENGINEERING AND SCIENCE. vol. 29, no. 10, Oct ,,. 0.5 %.., cm mm FR4 (ε r =4.4)

... K-vision Fig.. K-vision camera tracking screen Drummond [3] 3. 3 (lines), (edge) 3. (target). (homography perspective transform) [4]. (drifting).

Transcription:

1 딥러닝기반성별및연령대 추정을통한맞춤형광고솔루션 20101588 조준희 20131461 신혜인

2 개요 연구배경 맞춤형광고의필요성 성별및연령별주요관심사에적합한광고의필요성증가 제한된환경에서개인정보획득의한계 맞춤형광고의어려움 영상정보기반개인정보추정 연구목표 딥러닝기반사용자맞춤형광고솔루션구현 얼굴영상을이용한성별및연령대추정 성별및연령대를통합네트워크로학습하여추정정확도향상 근실시간으로다수의사용자에대해추정

3 기존의연구 딥러닝기반나이추정 5 딥러닝사용전 딥러닝사용후 얼굴영상기반으로 4 성별, 나이추정 3 딥러닝사용시향상된성능 2 평균오차 ( 나이 ) 1 기존연구의한계 2 개의개별네트워크 ( 성별네트워크와나이네트워크 ) 를사용하여성별및나이추정 [1] 0 Huerta 등 [4] Guo 와 Mu [3] Yi 등 [2] Rothe 등 [1] [1] R. Rothe, R. Timofte, and L. V. Gool, DEX: Deep expectation of apparent age from a single image, in Proc. IEEE Int. Conf. Computer Vision, pp. 252 257, Santiago, Chile, Dec. 2015. [2] Yi. D, Lei. Z, and Li SZ, Age estimation by multi-scale convolutional network,, in Proc. Asian Conf. Computer Vision, pp. 144 158, Singapore, Nov. 2014. [3] G. Guo and G. Mu, Simultaneous dimensionality reduction and human age estimation via kernel partial least squares regression, in Proc. IEEE Int. Conf. Computer Vision and Pattern Recognition, pp. 657 664, Washington DC, USA, Jun. 2011. [4] I. Huerta, C. Fernández, and A. Prati, "Facial age estimation through the fusion of texture and local appearance descriptors," in Proc. European Conf. Computer Vision Workshop, pp. 667 681, Zurich, Switzerland, Sep. 2014.

4 시스템블록도및학습블록도 시스템블록도 I in 얼굴영상전처리 I f p v 성별및연령대추정 Deep Learning based Age and Gender Prediction 광고선택 학습과정

5 AlexNet 을이용한학습 (1/3) 2 개의네트워크를사용한학습 성별네트워크 연령대네트워크 통합네트워크를사용한학습 성별과연령대를통합네트워크로학습 연령대 남성 Multi-label binary encoding [5]: 14 개의클래스 성별및연령대네트워크 7 개의연령대클래스 출처 : 2015 소비자행태조사보고서 라벨 여성 0~12 0 7 13~18 1 8 19~29 2 9 30~39 3 10 40~49 4 11 50~64 5 12 65~ 6 13 2 개의성별클래스 [5] Li et al., Deep: Learning deep binary encoding for multi-label classification, in Proc. IEEE Int. Conf. Computer Vision and Pattern Recognition Workshop, pp. 39 46, Las Vegas, NV, USA, Jun. 2016.

6 AlexNet 을이용한학습 (2/3) AlexNet vs. VGG-16 ImageNet dataset (1000 개클래스 ) [8] AlexNet 에비해 VGG-16 은깊은구조로보다다양한특징추출 물체분류의경우다양한특징추출에따라성능향상 AlexNet 사용이유 AlexNet [6] VGG-16 [7] 파라미터수 ( 억 ) 0.61 1.38 연산량 ( 억 ) 7.25 154.84 Top-1 에러 (%) 38.1 23.7 파라미터수증가에따른메모리및연산량증가 얼굴영상을통한성별및연령대추정은많은특징을필요로하지않아얕은구조의 AlexNet 으로도우수한성능을보임 [6] A. Krizhevsky, I. Sutskever, and G. E. Hinton, ImageNet classification with deep convolutional neural networks, in Proc. IEEE Int. Conf. Advances in Neural Information Processing Systems, pp. 1106 1114, Lake Tahoe, NV, USA, Dec. 2012. [7] S. Karen and Z. Andrew, Very deep convolutional networks for large-scale image recognition, in Proc. Int. Conf. Learning Representation, pp. 1 13, San Diego, CA, USA, May 2015. [8] Russakovsky et al., ImageNet large scale visual recognition challenge, Int. Journ. Computer Vision, vol. 115, no. 3, pp. 211 252, Dec. 2015.

7 AlexNet 을이용한학습 (3/3) Fine tunning [9] ImageNet dataset 을이용한 AlexNet 학습모델사용 [6] Convolutional layer 파라미터유지, fully connected layer 파라미터초기화후학습진행 어플리케이션에적합한학습모델생성 IMDB-WIKI dataset 정제 [1] 얼굴이아닌영상이포함되어학습성능저하 정렬된얼굴영상취득 우수한성능의얼굴검출및 head pose 추정알고리즘 [10] 사용 [1] R. Rothe, R. Timofte, and L. V. Gool, DEX: Deep expectation of apparent age from a single image, in Proc. IEEE Int. Conf. Computer Vision, pp. 252 257, Santiago, Chile, Dec. 2015. [6] A. Krizhevsky, I. Sutskever, and G. E. Hinton, ImageNet classification with deep convolutional neural networks, in Proc. IEEE Int. Conf. Advances in Neural Information Processing Systems, pp. 1106 1114, Lake Tahoe, NV, USA, Dec. 2012. [9] S. J. Pan and Q. Yang, A survey on transfer learning, IEEE Trans. Knowledge and Data Engineering, vol. 22, no. 10, pp. 1345 1359, Oct. 2010. [10] A. Asthana, S. Zafeiriou, S. Cheng, and M. Pantic, Incremental face alignment in the wild, in Proc. IEEE Int. Conf. Computer Vision and Pattern Recognition, pp. 1859 1866, Columbus, OH, USA, Jun. 2014.

8 얼굴영상전처리 웹캠영상속얼굴영상처리 얼굴검출 얼굴영역추출 얼굴정렬 입력영상 정렬된얼굴영상 영상에서얼굴검출후정렬 Haar-like face detection [12] 얼굴검출 Head pose estimation [11] 검출된얼굴에서 head pose 추정후정렬된얼굴영상취득 [11] A. Asthana, S. Zafeiriou, S. Cheng and M. Pantic, Incremental face alignment in the wild, in Proc. IEEE Int. Conf. Computer Vision and Pattern Recognition, pp. 1859 1866, Columbus, OH, USA, Jun. 2014. [12] P. Viola and M. J. Jones, Rapid object detection using a boosted cascade of simple features, in Proc. IEEE Int. Conf. Computer Vision and Pattern Recognition, pp. 511 518, Kauai, HI, USA, Dec. 2001.

9 성별및연령대추정 AlexNet 모델이용 얼굴검출후정렬된얼굴영상을입력 사전학습모델기반연령대와성별추정 학습된모델의커널과의콘볼루션연산을통한특징추출 14개클래스중가장높은점수의클래스출력

10 광고선택 성별, 연령대별관심사 80 70 60 50 40 30 20 10 0 13~18 세남성 19~29 세남성 30~39 세남성 40~49 세남성 50~64 세남성 13~18 세여성 19~29 세여성 30~39 세여성 40~49 세여성 50~64 세여성 음료화장품자동차용품출처 : 2015 소비자행태조사보고서 성별및연령대에따라관심있는광고차이 다수인원검출시비율에따른광고출력

시스템구현 얼굴영상전처리 웹캠영상속사람얼굴검출후기울어진얼굴정렬 딥러닝기반성별및연령대추정 사전학습모델 (AlexNet) 을 fine tunning하여사용자의성별및연령대를통합네트워크로학습 맞춤형광고출력 사용자의추정된성별및연령대를바탕으로선별적광고 11 이름 역할 조준희 AlexNet을이용한학습진행 신혜인 GUI 및시스템구현 7월 8월 9월 10월 11월프로젝트주제선정및자료조사얼굴검출딥러닝기반성별및연령대추정맞춤형광고선별최종시연

12 실험결과 (1/3) 개발환경 GPU: NVIDIA GeForce GTX 970 (Memory: 4 GB) CPU: Intel i5-4670 3.4 GHz 학습소요시간 : 약 4일 결과비교 IMDB-WIKI dataset [1] 네트워크개수에따른 AlexNet 학습모델성별및연령대추정정확도비교 정확도 (%) 통합네트워크 73.15 2 개의네트워크 71.05 통합네트워크사용시성별및연령대추정정확도향상 [1] R. Rothe, R. Timofte, and L. V. Gool, DEX: Deep expectation of apparent age from a single image, in Proc. IEEE Int. Conf. Computer Vision, pp. 252 257, Santiago, Chile, Dec. 2015.

13 실험결과 (2/3) 전체시스템결과 다양한환경에서정확한성별및연령대추정 (26 세남성 ) 선글라스를안경쓴경우안경벗은경우모자를쓴경우밝은조명어두운조명쓴경우 표정변화에따른추정연령대오차발생 주름으로인해실제나이에비해높은연령대로추정 인상쓴얼굴

14 실험결과 (3/3) 전체시스템결과 다수사용자의나이와성별에따른광고재생 성별, 연령대추정 광고재생비율 19~29 세여성광고재생 33% 67% 19~29 세남성광고재생 추정인원구성 : 19~29 세여성 1 명 19~29 세남성 2 명 인원비율에따른광고재생

15 결론및추후과제 결론 딥러닝기반얼굴영상을통한성별및연령대추정 통합네트워크를사용하여성별및연령대추정정확도향상 다수사용자의성별및연령대추정구현 실시간에가까운사용자맞춤형광고솔루션구현 추정된성별및연령대기반사용자맞춤형광고출력 추후과제 성별및연령대추정정확도향상 표정변화에도강건한성별및연령대추정 연령대및성별외의새로운요소들 ( 표정, 복장등 ) 을이용한맞춤형광고솔루션구현