bag-of-words bag-of-keypoints Web bagof-keypoints Nearest Neighbor SVM Nearest Neighbor SIFT Nearest Neighbor bag-of-keypoints Nearest Neighbor SVM 84

Similar documents
IPSJ SIG Technical Report Vol.2010-CVIM-170 No /1/ Visual Recognition of Wire Harnesses for Automated Wiring Masaki Yoneda, 1 Ta

IPSJ SIG Technical Report Vol.2011-CVIM-177 No /5/ TRECVID2010 SURF Bag-of-Features 1 TRECVID SVM 700% MKL-SVM 883% TRECVID2010 MKL-SVM A

(MIRU2009) cuboid cuboid SURF 6 85% Web. Web Abstract Extracting Spatio-te

Google Goggles [1] Google Goggles Android iphone web Google Goggles Lee [2] Lee iphone () [3] [4] [5] [6] [7] [8] [9] [10] :

LBP 2 LBP 2. 2 Local Binary Pattern Local Binary pattern(lbp) [6] R

(MIRU2008) HOG Histograms of Oriented Gradients (HOG)

情報処理学会研究報告 IPSJ SIG Technical Report Vol.2013-CVIM-186 No /3/15 EMD 1,a) SIFT. SIFT Bag-of-keypoints. SIFT SIFT.. Earth Mover s Distance

IS1-09 第 回画像センシングシンポジウム, 横浜,14 年 6 月 2 Hough Forest Hough Forest[6] Random Forest( [5]) Random Forest Hough Forest Hough Forest 2.1 Hough Forest 1 2.2

(MIRU2010) Geometric Context Randomized Trees Geometric Context Rand

Duplicate Near Duplicate Intact Partial Copy Original Image Near Partial Copy Near Partial Copy with a background (a) (b) 2 1 [6] SIFT SIFT SIF

Microsoft PowerPoint - SSII_harada pptx

2. 30 Visual Words TF-IDF Lowe [4] Scale-Invarient Feature Transform (SIFT) Bay [1] Speeded Up Robust Features (SURF) SIFT 128 SURF 64 Visual Words Ni

THE INSTITUTE OF ELECTRONICS, INFORMATION AND COMMUNICATION ENGINEERS TECHNICAL REPORT OF IEICE. TRECVID2012 Instance Search {sak

2.2 6).,.,.,. Yang, 7).,,.,,. 2.3 SIFT SIFT (Scale-Invariant Feature Transform) 8).,. SIFT,,. SIFT, Mean-Shift 9)., SIFT,., SIFT,. 3.,.,,,,,.,,,., 1,

本文6(599) (Page 601)

1 (PCA) 3 2 P.Viola 2) Viola AdaBoost 1 Viola OpenCV 3) Web OpenCV T.L.Berg PCA kpca LDA k-means 4) Berg 95% Berg Web k-means k-means

x T = (x 1,, x M ) x T x M K C 1,, C K 22 x w y 1: 2 2

IPSJ SIG Technical Report Vol.2012-CG-149 No.13 Vol.2012-CVIM-184 No /12/4 3 1,a) ( ) DB 3D DB 2D,,,, PnP(Perspective n-point), Ransa

untitled

4. C i k = 2 k-means C 1 i, C 2 i 5. C i x i p [ f(θ i ; x) = (2π) p 2 Vi 1 2 exp (x µ ] i) t V 1 i (x µ i ) 2 BIC BIC = 2 log L( ˆθ i ; x i C i ) + q

Convolutional Neural Network A Graduation Thesis of College of Engineering, Chubu University Investigation of feature extraction by Convolution

yoo_graduation_thesis.dvi

(b) BoF codeword codeword BoF (c) BoF Fergus Weber [11] Weber [12] Weber Fergus BoF (b) Fergus [13] Fergus 2. Fergus 2. 1 Fergus [3]

3 2 2 (1) (2) (3) (4) 4 4 AdaBoost 2. [11] Onishi&Yoda [8] Iwashita&Stoica [5] 4 [3] 3. 3 (1) (2) (3)

(VKIR) VKIR VKIR DCT (R) (G) (B) Ward DCT i

DEIM Forum 2010 A Web Abstract Classification Method for Revie

一般社団法人電子情報通信学会 THE INSTITUTE OF ELECTRONICS, INFORMATION AND COMMUNICATION ENGINEERS THE INSTITUTE OF ELECTRONICS, INFORMATION AND COMMUNICATION ENGIN

[1] SBS [2] SBS Random Forests[3] Random Forests ii

% 2 3 [1] Semantic Texton Forests STFs [1] ( ) STFs STFs ColorSelf-Simlarity CSS [2] ii

IPSJ SIG Technical Report Vol.2013-CVIM-187 No /5/30 1,a) 1,b), 1,,,,,,, (DNN),,,, 2 (CNN),, 1.,,,,,,,,,,,,,,,,,, [1], [6], [7], [12], [13]., [

IPSJ SIG Technical Report Vol.2010-CVIM-171 No /3/19 1. Web 1 1 Web Web Web Multiple Kernel Learning(MKL) Web ( ) % MKL 68.8% Extractin

28 TCG SURF Card recognition using SURF in TCG play video

IPSJ SIG Technical Report iphone iphone,,., OpenGl ES 2.0 GLSL(OpenGL Shading Language), iphone GPGPU(General-Purpose Computing on Graphics Proc

& 3 3 ' ' (., (Pixel), (Light Intensity) (Random Variable). (Joint Probability). V., V = {,,, V }. i x i x = (x, x,, x V ) T. x i i (State Variable),

A Survey on Image Recognition Using Geo-tag Information

3: 2: 2. 2 Semi-supervised learning Semi-supervised learning [5,6] Semi-supervised learning Self-training [13] [14] Self-training Self-training Semi-s

THE INSTITUTE OF ELECTRONICS, INFORMATION AND COMMUNICATION ENGINEERS TECHNICAL REPORT OF IEICE.,,

光学

24 Region-Based Image Retrieval using Fuzzy Clustering

,,, Twitter,,, ( ), 2. [1],,, ( ),,.,, Sungho Jeon [2], Twitter 4 URL, SVM,, , , URL F., SVM,, 4 SVM, F,.,,,,, [3], 1 [2] Step Entered

untitled

一般画像認識のための単語概念の視覚性の分析

Personal Software Process /Team Software Process (PSP/TSP) Watts Humphrey PSP/TSP PSP/TSP TaskPit TaskPit PSP/TSP TaskiPit PSP/TSP 5 5

IPSJ SIG Technical Report Vol.2009-CVIM-167 No /6/10 Real AdaBoost HOG 1 1 1, 2 1 Real AdaBoost HOG HOG Real AdaBoost HOG A Method for Reducing

Microsoft PowerPoint - cvim_harada pptx

IPSJ SIG Technical Report 1,a) 1,b) 1,c) 1,d) 2,e) 2,f) 2,g) 1. [1] [2] 2 [3] Osaka Prefecture University 1 1, Gakuencho, Naka, Sakai,

12_39.dvi

Sobel Canny i

The 18th Game Programming Workshop ,a) 1,b) 1,c) 2,d) 1,e) 1,f) Adapting One-Player Mahjong Players to Four-Player Mahjong

kut-paper-template.dvi

ISCO自動コーディングシステムの分類精度向上に向けて―SSM およびJGSS データセットによる実験の結果―

SICE東北支部研究集会資料(2013年)

Haiku Generation Based on Motif Images Using Deep Learning Koki Yoneda 1 Soichiro Yokoyama 2 Tomohisa Yamashita 2 Hidenori Kawamura Scho

No. 3 Oct The person to the left of the stool carried the traffic-cone towards the trash-can. α α β α α β α α β α Track2 Track3 Track1 Track0 1

1., 1 COOKPAD 2, Web.,,,,,,.,, [1]., 5.,, [2].,,.,.,, 5, [3].,,,.,, [4], 33,.,,.,,.. 2.,, 3.., 4., 5., ,. 1.,,., 2.,. 1,,

Microsoft Word - toyoshima-deim2011.doc

画像認識性能を改善する高精度な特徴量抽出手法の検討 A Study on Feature-Extraction Methods for Improvement of Image-Recognition Performance 井上俊明 Toshiaki Inoue 要旨 各種のカメラ搭載機器の急速な

THE INSTITUTE OF ELECTRONICS, INFORMATION AND COMMUNICATION ENGINEERS TECHNICAL REPORT OF IEICE. Wang Jiani {jwang,mnod

IPSJ SIG Technical Report Vol.2011-EC-19 No /3/ ,.,., Peg-Scope Viewer,,.,,,,. Utilization of Watching Logs for Support of Multi-

IPSJ SIG Technical Report GPS LAN GPS LAN GPS LAN Location Identification by sphere image and hybrid sensing Takayuki Katahira, 1 Yoshio Iwai 1

…p…^†[…fiflF”¯ Pattern Recognition

(a) 1 (b) 3. Gilbert Pernicka[2] Treibitz Schechner[3] Narasimhan [4] Kim [5] Nayar [6] [7][8][9] 2. X X X [10] [11] L L t L s L = L t + L s

Honda 3) Fujii 4) 5) Agrawala 6) Osaragi 7) Grabler 8) Web Web c 2010 Information Processing Society of Japan

28 Horizontal angle correction using straight line detection in an equirectangular image

80 Sep CBIR 6),7) WWW WWW Image Collector Image Collector (1) (2) 1 WWW 2 CBIR WWW WWW WWW CBIR example-based generic object recognition 8),9) W

BDH Cao BDH BDH Cao Cao Cao BDH ()*$ +,-+.)*$!%&'$!"#$ 2. 1 Weng [4] Metric Learning Weng DB DB Yang [5] John [6] Sparse Coding sparse coding DB [7] K

1 Web [2] Web [3] [4] [5], [6] [7] [8] S.W. [9] 3. MeetingShelf Web MeetingShelf MeetingShelf (1) (2) (3) (4) (5) Web MeetingShelf

IPSJ SIG Technical Report Vol.2017-CVIM-205 No /1/ Content-based Image Retrieval(CBIR) CBIR RANSAC (Local feature hashing) 1000 A geo


IPSJ SIG Technical Report Vol.2017-MUS-116 No /8/24 MachineDancing: 1,a) 1,b) 3 MachineDancing MachineDancing MachineDancing 1 MachineDan

(a) (b) 2 2 (Bosch, IR Illuminator 850 nm, UFLED30-8BD) ( 7[m] 6[m]) 3 (PointGrey Research Inc.Grasshopper2 M/C) Hz (a) (b

Automatic Collection of Web Video Shots Corresponding to Specific Actions using Web Images

Drive-by-Download JavaScript

Twitter

Computational Semantics 1 category specificity Warrington (1975); Warrington & Shallice (1979, 1984) 2 basic level superiority 3 super-ordinate catego

WISS Woodman Labs GoPro 1 [5, 3, 2] Copyright is held by the author(s). 1 GoPro GoPro 2 6 GoPro RICOH THETA 3 Kodak P

paper.dvi


HOG HOG LBP LBP 4) LBP LBP Wang LBP HOG LBP 5) LBP LBP 1 r n 1 n, 1

29 jjencode JavaScript

Silhouette on Image Object Silhouette on Images Object 1 Fig. 1 Visual cone Fig. 2 2 Volume intersection method Fig. 3 3 Background subtraction Fig. 4

Fig. 2 28th Ryuou Tournament, Match 5, 59th move. The last move is Black s Rx5f. 1 Tic-Tac-Toe Fig. 1 AsearchtreeofTic-Tac-Toe. [2] [3], [4]

Optical Flow t t + δt 1 Motion Field 3 3 1) 2) 3) Lucas-Kanade 4) 1 t (x, y) I(x, y, t)

[2] OCR [3], [4] [5] [6] [4], [7] [8], [9] 1 [10] Fig. 1 Current arrangement and size of ruby. 2 Fig. 2 Typography combined with printing

,,.,.,,.,.,.,.,,.,..,,,, i

..,,,, , ( ) 3.,., 3.,., 500, 233.,, 3,,.,, i

Microsoft PowerPoint - 03Weka.ppt

2 ( ) i

Microsoft PowerPoint - pr_12_template-bs.pptx

1 Fig. 1 Extraction of motion,.,,, 4,,, 3., 1, 2. 2.,. CHLAC,. 2.1,. (256 ).,., CHLAC. CHLAC, HLAC. 2.3 (HLAC ) r,.,. HLAC. N. 2 HLAC Fig. 2

Computer Security Symposium October ,a) 1,b) Microsoft Kinect Kinect, Takafumi Mori 1,a) Hiroaki Kikuchi 1,b) [1] 1 Meiji U

IPSJ SIG Technical Report Vol.2013-GN-86 No.35 Vol.2013-CDS-6 No /1/17 1,a) 2,b) (1) (2) (3) Development of Mobile Multilingual Medical

1 Kinect for Windows M = [X Y Z] T M = [X Y Z ] T f (u,v) w 3.2 [11] [7] u = f X +u Z 0 δ u (X,Y,Z ) (5) v = f Y Z +v 0 δ v (X,Y,Z ) (6) w = Z +

Twitter‡Ì”À‰µ…c…C†[…g‡ðŠŸŠp‡µ‡½…^…C…•…›…C…fi‘ã‡Ì…l…^…o…„‘îŁñ„�™m

aca-mk23.dvi

IPSJ SIG Technical Report Vol.2010-MPS-77 No /3/5 VR SIFT Virtual View Generation in Hallway of Cybercity Buildings from Video Sequen

SICE東北支部研究集会資料（2017年）

24 SPAM Performance Comparison of Machine Learning Algorithms for SPAM Discrimination

untitled

Vol1-CVIM-172 No.7 21/5/ Shan 1) 2 2)3) Yuan 4) Ancuti 5) Agrawal 6) 2.4 Ben-Ezra 7)8) Raskar 9) Image domain Blur image l PSF b / = F(

Transcription:

Bag-of-Keypoints Web G.Csurka bag-of-keypoints Web Bag-of-keypoints SVM 5.% Web Image Classification with Bag-of-Keypoints Taichi joutou and Keiji yanai Recently, need for generic image recognition is getting larger due to the explosive increase of digital images. Then, we have performed classification expreiments for general images gathered from Web employing the bag-of-keypoints method proposed by G.Csurka et al. In the experiments, we have obtained the 5.% classification rate for object classes... World Wide Web Flickr ) Web Web Department of Computer Science, The University of Electro-Communications Corel Corel Image Gallery. G. Csurka bag-of-keypoints ) Web.3

bag-of-words bag-of-keypoints Web bagof-keypoints Nearest Neighbor SVM Nearest Neighbor SIFT Nearest Neighbor bag-of-keypoints Nearest Neighbor SVM 84% 5%. SIFT SIFT David Lowe 3) 4) SIFT Sivic, J. 5) 4) G. Csurka ) bag-of-keypoints SIFT Naive Bayes SVM 7 SIFT bag-of-keypoints Sivic, J. G. Csurka SIFT bag-of-keypoints 3. Web SIFT bag-of-keypoints ) ( ) k-means visual words Nearest Neighbor SVM Nearest Neighbor SIFT Nearest Neighbor bag-of-keypoints Nearest Neighbor SVM 5 bag-of-keypoints Web

4. 4. 4.. 3 3 4.. (SIFT) SIFT(Scale Invariant Feature Transformation) 3) David Lowe SIFT DOG(Difference of Gaussian) 8 SIFT SIFT ( ) 8 (4 4 8) 3D SIFT C++ SIFT++ 6) pgm SIFT (8 ) SIFT++ C++ 4. Bag-of-keypoints Bag-of-keypoints ) visual words code book visual word (bag) Bag-of-keypoints ( ) ( ) codebook ( 3 ) codebook ( 4 ) 4.. Visual words visual word SIFT (4.. ) code book visual words( ) visual words visual words visual words 4... k-means kmeans k k code book k 3 5 8 5 4.. Code book code book (bag-of-keypoints) code book () ( 3) code book visual word code 3

book k k 3 4.3 / SVM multiclass (Multi-Class Support Vector Machine) 7) 4.3. 4.3.. M.J.Swain 8) I j j M j j S IM n S IM = min(i j, M j) j= 4.3.. (Nearest Neighbor) () 4.3. SVM(Support Vector Machine) SVM 9) 96 Vapnik Optimal Separating Hyperplane SVM SVM SVM light) SVM multiclass (Multi-Class support Vector Machine) 7) SVM 4.3.. Φ (Φ(x) Φ(x )) K(x, x ) linear radial basis function linear( ) y = sign(w T x h) radial basis function() K(x, x ) = exp( x x ) σ radial basis function SVM Radial Basis Function(RBF) 5. 5. ( ) Web ( ) ( ) ( ) 5 ) ( 4) 4

5 5 ) ( ) 8 9 6 3 5 6 4 5 () () 5. ( 4) Nearest Neighbor bag-of-keypoints Nearest Neighbor 3 bag-of-keypoints SVM( linear) 4 bag-of-keypoints SVM( radial basis function) bag-of-keypoints code book 3 5 8 5.3 5-fold cross validation 5 (( )/( )) (( )/( )) F- (F-measure) (confusion matrix) 5.3. F- (F-measure) F- (recall) (precision) () F = + recall precision 5.3. (confusion matrix) M ij M ij = {I k C j : h(i k ) = i} C j i, j {,..., N c}(n c ) C j j h(i k ) I k 5.4 5.4. 4 3 4 5

図4 種類の画像のサンプル (左上から グ ローブ ) す 手法 では code book のサイズ別に表記した 次に 手法 の分類器を Nearest Neighbor から SVM に変えて分類実験した結果を表 5 と表 6 で示す SVM はカーネルを 通り試したため カーネル別に 分けてから code book サイズで分けて表記した また クラス別にそれぞれの手法の F-値を計算し たものを表 5 で示す ただし bag-of-keypoints を用 いた手法においては 結果がよかった code book サ イズの実験結果のみを示す 最後に 種類の F-値の平均が最も良かった カーネルに radial basis function を用いた bag-ofkeypoints+svm の手法 (手法 4) で code book サイ ズを 8 としたときの結果を混合行列にして表 7 で 示す 図 5 種類の F-値とその平均 5.4. 種類分類実験の結果 種類分類の結果より SVM を用いた手法におい てカーネルの違いによる大きな結果の違いがなかった ため 学習時間が圧倒的に短い線形 (linear) カーネル による SVM を 種類分類において採用する 種類分類の結果を 種類の平均再現率と平均 適合率を用いて表 8 で示す さらに F-値で評価した 結果の上位 種類と 下位 種類を表 9 で示す 結 果が特に良かったもの 悪かったものを図 6 で示す 表 3 カラー+NN と bag-of-keypoints+nn の再現率 平均 カラー ヒストグラム.4.4.68.68.6.48.8.6.48.5.5 bag-of-keypoints(code book size(c)) 3 5 8.74.9.86.78.96.86.74.8.7.8.9.9.98.96.98.54.74.66.6.56.9.86.9..94.34.34.38..34.6.8.7.54.9.7.58.64.6.9.84.9.88.9.9.94.98.88.9.8.74.78.77.69.8 表8 種類の分類結果 平均再現率.5 平均適合率.489 表 4 カラー+NN と bag-of-keypoints+nn の適合率 平均 カラー ヒストグラム.46.5.83.54.36.48.43.6.5.58.53 表9 bag-of-keypoints(code book size(c)) 3 5 8.59.6.7.46.7.76.69.87.79.83.75.8.66.7.6.96.78.8.6.78.86.75.87.87.94.43.74.68.35.78.76..75.6.89.84.8.98...87.9.8.77.94.8.79.79.77.8.76.79.79.69.83 種類の F-値の TOP と WORST 自転車 ネックレス 阿波踊り 花火大会 スパゲッティ 温泉 剣道 ねぶた 石垣 スーツ 6 TOP.87.78.77.76.76.76.76.75.74.73 橋 港 神社 猫 相撲 アスパラガス 川 うなぎ トカゲ ツツジ WORST.3.5.5.8..4.7.8.8.9

表 5 bag-of-keypoints+svm での再現率 カーネル 平均.7.8.5.9.96..76.6.9.9.7 linear code book size 3 5 8.6.6.34.56.8.9.6.68.94.94.8.8.94.96...44.54.76.96.84.88.96..8.83.9.6.9.84.68.79.8.9.96.46.88..4.9.8.98.94.79 radial basis function code book size 3 5 8.6.64.34.56.8.86.66.7.94.94.84.86.94.96...5.56.78.96.84.8.9.98.84.84.9.64.9.8.69.8.8.8.74.58.8.98.6.9.6.9.9.73.9.98.48.84...9.8.98.96.8 表 6 bag-of-keypoints+svm での適合率 カーネル 平均.59.89..69.6..9.7.84.63.68 linear code book size 3 5 8.63.97.85.7.98.98.66.89.96.68.87.75.65.59.68.3.8.8.93.73.84.68.67.68.7.9.87.8.85.9.6.74.83.88...7.63..83.69.94.77.77.56.9.97.74.77.8.8.8.8.6.7 radial basis function code book size 3 5 8.55.94.74.68.98.98.67.88.96.7.88.7.64.6.7.4.87.88.93.73.8.75.73.7.7.89.85.8.87.87.67.84.84.85...7.65.35.79.78.94.8.79 表 7 手法 4(code book size=8) の混合行列 True class 7 5 43 47 4 43 7 5 5 8 8 6 3 4 49 5 3 45 3 4 図 6 左図が結果がよかった例 右図が結果が悪かった例 5.5 考 察 5.5. 種類の分類 まず 手法 と手法 の結果より カラーヒストグ ラムと SIFT 特徴 (bag-of-keypoints) の特徴量につい て考察してみる 表 3 表 4 を見ると 再現率 適合率 とも SIFT 特徴の方が 圧倒的に認識率が優れている 特に の画像は図 4 を見るとわかるように には 角や丸みを帯びたものなど他の種類にない物体 の形状の特徴を持っている 逆に 色の視点から見て みると 背景に建物や空が写っていて 色だけでは判 断しにくいことがわかる このことにより カラーヒ ストグラムより SIFT 特徴の方が特徴量として有効だ と示される 次に 手法 と手法 3 の結果より Nearest Neighbor と SVM の分類器について考察してみる 表 3 表 4 より 手法 の再現率と適合率の平均は.76 と.77 であり 手法 3 は表 5 表 6 より 線形カーネルを用 いた SVM のものは.76 と.73 で若干手法 の方 7

SVM.77.77 3 4 F- ( 5) SVM bag-of-keypoints 8 5.5. ( 8) 5% %.94.8..4 9 4 6. Web 84% 5% Nearest Neighbor visual words bag-of-keypoints O.Maron ) Multiple Instance Learning(MIL) MIL () O.Maron Diverse Density MIL () Chen, Y. ) MIL Chen, Y. ) Flickr: http://www.flickr.com/. ) Csurka, G., Bray, C., Dance, C. and Fan, L.: Visual categorization with bags of keypoints, Workshop on Statistical Learning in Computer Vision,European Conference on Computer Vision, pp. (4). 3) Lowe, D.: Distinctive Image Features from Scale-Invariant Keypoints, International Journal of Computer Vision, Vol.6, No., pp.9 (4). 4) Brown, M. and Lowe, D.: Recognising panoramas, Proc. The International Conference on Computer Vision, pp.8 5 (3). 5) Sivic, J. and Zisserman, A.: Video Google: A Text Retrieval Approach to Object Matching in Videos, Proceedings of the International Conference on Computer Vision, Vol., pp. 47 477 (3). 6) Vedaldi, A.: SIFT++, http://vision.ucla. edu/ vedaldi/code/siftpp/siftpp.html. 7) Joachims, T.: SVM multiclass, http://www.cs. cornell.edu/people/tj/svm light/svm multiclass. html. 8) M.J.Swain and D.H.Ballard: Color Indexing, International Journal of Computer Vision, Vol.7, No., pp. 3 (99). 9) (3). ) Joachims, T.: SVM light, http://www.cs. cornell.edu/people/tj/svm light/index.html. ) Maron, O. and Ratan, A.: Multiple-instance learning for natural scene classification, The Fifteenth International Conference on Machine Learning, pp.34 349 (998). ) Chen, Y., Bi, J. and Wang, J.: MILES: Multiple-Instance Learning via Embedded Instance Selection, IEEE Transaction on Pattern Analalysis and Machine Intelligence, Vol. 8, No., pp.93 947 (6). 8

2024 © docsplayer.net プライバシー | 利用規約 | フィードバック