{"id":2537,"date":"2011-12-06T14:48:19","date_gmt":"2011-12-06T05:48:19","guid":{"rendered":"http:\/\/www.moonmile.net\/blog\/archives\/2537"},"modified":"2011-12-07T12:09:20","modified_gmt":"2011-12-07T03:09:20","slug":"opencv-%e3%81%a7%e6%a9%9f%e6%a2%b0%e5%ad%a6%e7%bf%92%e3%82%92%e8%a9%a6%e3%81%97%e3%81%a6%e3%81%bf%e3%82%8b-%e3%81%a8%e3%81%be%e3%81%a0%e7%b5%82%e3%82%8f%e3%82%89%e3%81%9a-2","status":"publish","type":"post","link":"http:\/\/www.moonmile.net\/blog\/archives\/2537","title":{"rendered":"OpenCV \u3067\u6a5f\u68b0\u5b66\u7fd2\u3092\u8a66\u3057\u3066\u307f\u308b…\u3068\u307e\u3060\u7d42\u308f\u3089\u305a"},"content":{"rendered":"

\u30c6\u30f3\u30d7\u30ec\u30fc\u30c8\u30de\u30c3\u30c1\u30f3\u30b0\u3067\u753b\u50cf\u304b\u3089\u53d6\u308a\u51fa\u3059\u306b\u306f\u7121\u7406\u304c\u3042\u308a\u305d\u3046\u306a\u306e\u3067\u3001\u3084\u306f\u308a\u3001\u72ec\u81ea\u306a\u30aa\u30d6\u30b8\u30a7\u30af\u30c8\u691c\u51fa\u5668\u3092\u4f5c\u3063\u3066\u307f\u306a\u3044\u3068\u3060\u3081\u304b\u3001\u3068\u601d\u3044\u3001<\/p>\n

OpenCV\u3067\u5b66\u3076\u753b\u50cf\u8a8d\u8b58\uff1a\u7b2c4\u56de\u3000\u30aa\u30d6\u30b8\u30a7\u30af\u30c8\u691c\u51fa\u5668\u306e\u4f5c\u6210\u65b9\u6cd5\uff5cgihyo.jp \u2026 \u6280\u8853\u8a55\u8ad6\u793e
\nhttp:\/\/gihyo.jp\/dev\/feature\/01\/opencv\/0004<\/a><\/p>\n

\u3092\u8aad\u3093\u3067\u3001\u8a66\u3057\u306b\u3001AdaBoost \u3092\u4f7f\u3063\u3066\u30ab\u30b9\u30b1\u30fc\u30c9\u3092\u4f5c\u308d\u3046\u3068\u601d\u3063\u305f\u306e\u3060\u304c\u3001\u3044\u3084\u3042\u3001\u3061\u3087\u3063\u3068\u6642\u9593\u304c\u639b\u304b\u308a\u3059\u304e\u308b\u3002<\/p>\n

\u25a0\u5b66\u7fd2\u7528\u306e\u6b63\u89e3\u30d5\u30a1\u30a4\u30eb\u306e\u4f5c\u6210<\/p>\n

\r\nC:\\OpenCV2.3\\build\\bin\\opencv_createsamples.exe ^\r\n -img images\\koma01.png ^\r\n -vec koma01.vec ^\r\n -num 1000 ^\r\n -bg NG.txt ^\r\n -w 45 -h 45 ^\r\n -show\r\n\r\n<\/pre>\n
\u25a0\u30aa\u30d6\u30b8\u30a7\u30af\u30c8\u306e\u5b66\u7fd2<\/p>\n
\r\nC:\\OpenCV2.3\\build\\bin\\opencv_haartraining.exe ^\r\n -data koma01 ^\r\n -vec  koma01.vec ^\r\n -bg NG.txt ^\r\n -npos 1000 ^\r\n -nneg 467 ^\r\n -w 45 -h 45 ^\r\n -mem 200 ^\r\n\r\n<\/pre>\n
\u306a\u611f\u3058\u3067\u52d5\u304b\u3057\u3066\u3044\u307e\u3059\u304c\u30014\u6642\u9593\u307b\u3069\u7d4c\u3063\u3066\u3082\u7d42\u308f\u3089\u305a\u3002<\/p>\n
\u99d2\u3092\u691c\u51fa\u3057\u305f\u3044\u306e\u3067\u3001\u4eca\u56de\u306e\u5834\u5408\u306f\u6b63\u89e3\u753b\u50cf\u306f1\u3064\uff08\u30b2\u30fc\u30e0\u4e2d\u306b\u51fa\u3066\u304f\u308b\u753b\u50cf\uff09\u306b\u306a\u308b\u306e\u3067\u3059\u304c\u3001\u5149\u306e\u95a2\u4fc2\u3084\u753b\u9762\u3092\u6620\u3059\u95a2\u4fc2\u304b\u3089\u3044\u304f\u3064\u304b\u306e\u6b63\u89e3\u753b\u50cf\u3092\u7528\u610f\u3057\u307e\u3059\u3002\u305d\u306e\u3042\u305f\u308a\u306f\u3001opencv_createsamples \u3092\u4f7f\u3063\u3066\u30011000 \u679a\u306e\u753b\u50cf\u306b\u6c34\u5897\u3057\u3057\u307e\u3059\u3002\u30d5\u30a1\u30a4\u30eb\u306f\u5225\u3005\u306b\u3067\u304d\u305a\u306b\u3001koma01.vec \u306e\u3088\u3046\u306b1\u3064\u306e\u30d5\u30a1\u30a4\u30eb\u306b\u307e\u3068\u3081\u3089\u308c\u307e\u3059\u3002<\/p>\n
\u6700\u521d\u306f\u3001\u99d2\u306e\u753b\u50cf\u304c 90×90 \u3060\u3063\u305f\u306e\u3067\u3001\u305d\u306e\u307e\u307e\u6307\u5b9a\u3057\u305f\u306e\u3067\u3059\u304c\u3001\u3042\u3048\u306a\u304f\u30e1\u30e2\u30ea\u30fc\u30aa\u30fc\u30d0\u30fc\u3057\u3066\u3057\u307e\u3044 opencv_haartraining \u304c\u30c0\u30a6\u30f3\u3002\u30b5\u30a4\u30ba\u3092 45×45 \u306b\u3059\u308b\u3068\u52d5\u304f\u3088\u3046\u306b\u306a\u3063\u305f\u306e\u3067\u3001\u5bfe\u8c61\u753b\u50cf\u306f\u305d\u3053\u305d\u3053\u5c0f\u3055\u3044\u30b5\u30a4\u30ba\u306b\u3057\u306a\u3044\u3068\u99c4\u76ee\u306a\u306e\u304b\u3082\u3002<\/p>\n
\u4e0d\u6b63\u89e3\u306e\u753b\u50cf\u3092\u3069\u306e\u3088\u3046\u306b\u96c6\u3081\u308b\u306e\u304b\uff1f\u3068\u3082\u601d\u3063\u305f\u306e\u3067\u3059\u304c\u3001\u30de\u30c3\u30c1\u30f3\u30b0\u3057\u306a\u3051\u308c\u3070\u4f55\u3067\u3082\u3044\u3044\u308f\u3051\u3067\u3001<\/p>\n
Caltech101
\nhttp:\/\/www.vision.caltech.edu\/Image_Datasets\/Caltech101\/Caltech101.html<\/a><\/p>\n
\u306b\u3042\u308b\u9069\u5f53\u306a\u30d5\u30a9\u30eb\u30c0\uff08BACKGROUND_Google \u3068\u3044\u3046\u30e9\u30f3\u30c0\u30e0\u3063\u307d\u3044\u753b\u50cf\u30d5\u30a9\u30eb\u30c0\u304c\u3042\u308b\uff09\u3092\u4f7f\u3063\u3066\u3044\u307e\u3059\u3002<\/p>\n
\u3046\u307e\u304f\u5b9f\u884c\u3067\u304d\u308b\u3068\u3001\u4e0b\u8a18\u306e\u3088\u3046\u306b\u3001\uff08\u591a\u5206\uff09\u30c6\u30f3\u30d7\u30ec\u30fc\u30c8\u753b\u50cf\u3092\u5207\u308a\u66ff\u3048\u306a\u304c\u3089\u3001\u95be\u5024\u3092\u5207\u308a\u66ff\u3048\u306a\u304c\u3089\u5b66\u7fd2\u3057\u59cb\u3081\u308b\u308f\u3051\u3067\u3059\u304c…<\/p>\n
\r\nD:\\work\\OpenCV\\src\\PiyoDetect\\PiyoML\\data>opencv_haartraining.exe  -d\r\nxml  -vec  koma01.vec  -bg NG.txt  -npos 1000  -nneg 467  -w 45 -h 45\r\n\r\nData dir name: koma01.xml\r\nVec file name: koma01.vec\r\nBG  file name: NG.txt, is a vecfile: no\r\nNum pos: 1000\r\nNum neg: 467\r\nNum stages: 14\r\nNum splits: 1 (stump as weak classifier)\r\nMem: 200 MB\r\nSymmetric: TRUE\r\nMin hit rate: 0.995000\r\nMax false alarm rate: 0.500000\r\nWeight trimming: 0.950000\r\nEqual weights: FALSE\r\nMode: BASIC\r\nWidth: 45\r\nHeight: 45\r\nApplied boosting algorithm: GAB\r\nError (valid only for Discrete and Real AdaBoost): misclass\r\nMax number of splits in tree cascade: 0\r\nMin number of positive samples per cluster: 500\r\nRequired leaf false alarm rate: 6.10352e-005\r\n\r\nTree Classifier\r\nStage\r\n+---+\r\n|  0|\r\n+---+\r\n\r\nNumber of features used : 1007032\r\n\r\nParent node: NULL\r\n\r\n*** 1 cluster ***\r\nPOS: 1000 1000 1.000000\r\nNEG: 467 1\r\nBACKGROUND PROCESSING TIME: 0.22\r\nPrecalculation time: 0.05\r\n+----+----+-+---------+---------+---------+---------+\r\n|  N |%SMP|F|  ST.THR |    HR   |    FA   | EXP. ERR|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   1|100%|-|-0.763750| 1.000000| 1.000000| 0.123381|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   2|100%|+|-0.660565| 0.999000| 0.837259| 0.123381|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   3| 94%|-|-1.246627| 0.996000| 0.490364| 0.104294|\r\n+----+----+-+---------+---------+---------+---------+\r\nStage training time: 889.36\r\nNumber of used features: 3\r\n\r\nParent node: NULL\r\nChosen number of splits: 0\r\n\r\nTotal number of splits: 0\r\n\r\nTree Classifier\r\nStage\r\n+---+\r\n|  0|\r\n+---+\r\n\r\n   0\r\n\r\nParent node: 0\r\n\r\n*** 1 cluster ***\r\nPOS: 996 1000 0.996000\r\nNEG: 465 0.524831\r\nBACKGROUND PROCESSING TIME: 0.03\r\nPrecalculation time: 0.05\r\n+----+----+-+---------+---------+---------+---------+\r\n|  N |%SMP|F|  ST.THR |    HR   |    FA   | EXP. ERR|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   1|100%|-|-0.635385| 1.000000| 1.000000| 0.182752|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   2|100%|+|-0.333012| 0.996988| 0.544086| 0.149213|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   3|100%|-|-0.999921| 0.996988| 0.544086| 0.082820|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   4|100%|+|-0.692312| 0.996988| 0.389247| 0.063655|\r\n+----+----+-+---------+---------+---------+---------+\r\nStage training time: 1205.64\r\nNumber of used features: 4\r\n\r\nParent node: 0\r\nChosen number of splits: 0\r\n\r\nTotal number of splits: 0\r\n\r\nTree Classifier\r\nStage\r\n+---+---+\r\n|  0|  1|\r\n+---+---+\r\n\r\n   0---1\r\n\r\nParent node: 1\r\n\r\n*** 1 cluster ***\r\nPOS: 993 1000 0.993000\r\nNEG: 463 0.230923\r\nBACKGROUND PROCESSING TIME: 0.13\r\nPrecalculation time: 0.05\r\n+----+----+-+---------+---------+---------+---------+\r\n|  N |%SMP|F|  ST.THR |    HR   |    FA   | EXP. ERR|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   1|100%|-|-0.912574| 1.000000| 1.000000| 0.160027|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   2|100%|+|-1.372380| 1.000000| 1.000000| 0.298077|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   3|100%|-|-1.029186| 1.000000| 1.000000| 0.073489|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   4|100%|+|-1.168364| 0.996979| 0.796976| 0.073489|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   5| 98%|-|-0.856654| 0.996979| 0.455724| 0.049451|\r\n+----+----+-+---------+---------+---------+---------+\r\nStage training time: 1487.86\r\nNumber of used features: 5\r\n\r\nParent node: 1\r\nChosen number of splits: 0\r\n\r\nTotal number of splits: 0\r\n\r\nTree Classifier\r\nStage\r\n+---+---+---+\r\n|  0|  1|  2|\r\n+---+---+---+\r\n\r\n   0---1---2\r\n\r\nParent node: 2\r\n\r\n*** 1 cluster ***\r\nPOS: 990 1000 0.990000\r\nNEG: 462 0.1232\r\nBACKGROUND PROCESSING TIME: 0.53\r\nPrecalculation time: 0.05\r\n+----+----+-+---------+---------+---------+---------+\r\n|  N |%SMP|F|  ST.THR |    HR   |    FA   | EXP. ERR|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   1|100%|-|-0.606609| 1.000000| 1.000000| 0.199036|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   2|100%|+|-0.848618| 1.000000| 1.000000| 0.199036|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   3|100%|-|-0.527684| 0.996970| 0.614719| 0.152204|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   4|100%|+|-0.781167| 0.996970| 0.623377| 0.154959|\r\n+----+----+-+---------+---------+---------+---------+\r\n|   5| 79%|-|-1.579804| 0.997980| 0.632035| 0.112948|\r\n\r\n<\/pre>\n
\u305d\u3082\u305d\u3082\u306e\u76ee\u7684\u304c\u3001\u30b2\u30fc\u30e0\u753b\u9762\u304b\u3089\u65e2\u77e5\u306e\u99d2\uff08\u6642\u306b\u306f\u672a\u77e5\u306e\u99d2\uff1f\uff09\u3092\u898b\u3064\u3051\u51fa\u3057\u305f\u3044\u308f\u3051\u3067\u3001\u9854\u8a8d\u8b58\u307b\u3069\u6b63\u78ba\u3067\u306a\u304f\u3066\u3082\u3088\u3044\u3057\u3001\u3042\u308b\u7a0b\u5ea6\u306e\u5834\u6240\u3092\u78ba\u5b9a\u3057\u3066\u304b\u3089\u518d\u30de\u30c3\u30c1\u3055\u305b\u308b\u3068\u3044\u3046\u65b9\u5f0f\u304c\u3084\u3063\u3071\u308a\u3088\u3055\u305d\u3046\u304b\u3001\u3068\u601d\u3044\u76f4\u3057\u3066\u3044\u308b\u6b21\u7b2c\u3067\u3059\u3002<\/p>\n
OpenCV\u3067\u5b66\u3076\u753b\u50cf\u8a8d\u8b58\uff1a\u7b2c3\u56de\u3000\u30aa\u30d6\u30b8\u30a7\u30af\u30c8\u691c\u51fa\u3057\u3066\u307f\u3088\u3046\uff5cgihyo.jp \u2026 \u6280\u8853\u8a55\u8ad6\u793e
\nhttp:\/\/gihyo.jp\/dev\/feature\/01\/opencv\/0003?page=2<\/a><\/p>\n
\u306b\u3042\u308b\u3088\u3046\u306b\u3001\u4f55\u3082\u306a\u3044\u3068\u3053\u308d\u304b\u3089\u7279\u5b9a\u306e\u3082\u306e\u3092\u898b\u3064\u3051\u308b\uff08\u691c\u51fa\u3059\u308b\u7387\u304c\u4f4e\u3044\uff09\u5834\u5408\u306b\u306f\u3001\u3053\u306e\u3088\u3046\u306a\u5b66\u7fd2\u304c\u3088\u3044\u306e\u304b\u3082\u3057\u308c\u307e\u305b\u3093\u304c\u3001\u3042\u3089\u304b\u3058\u3081\u99d2\u304c\u3042\u308b\u3068\u308f\u304b\u3063\u3066\u3044\u308b\u30b2\u30fc\u30e0\u306e\u76e4\u4e0a\u304b\u3089\u63a2\u3059\u306b\u306f\u3001\u3082\u3063\u3068\u524d\u51e6\u7406\u3092\u3057\u3066\u7d5e\u308a\u8fbc\u3093\u3067\u3082\u3088\u3044\u304b\u306a\u3068\u3002\u30aa\u30d6\u30b8\u30a7\u30af\u30c8\u691c\u51fa\u5668\u306e\u4e2d\u3067 Ture\/False \u306e\u5224\u65ad\u3092\u3059\u308b\u3088\u308a\u3082\u3001\u305d\u306e\u524d\u51e6\u7406\u3068\u3057\u3066 Ture\/False \u3092\u5927\u96d1\u628a\u306b\u5224\u65ad\u3059\u308b\u30ed\u30b8\u30c3\u30af\u3092\u631f\u3080\u5fc5\u8981\u304c\u3042\u308a\u305d\u3046\u3067\u3059\u3002\u3044\u3061\u3044\u3061\u3001\u99d2\u306e\u5f62\u72b6\u304c\u5909\u308f\u308b\u305f\u3073\u306b\u3001\u5b66\u7fd2\u3092\u3055\u305b\u308b\u308f\u3051\u306b\u306f\u3044\u304b\u306a\u3044\u3057\u3001\u5b66\u7fd2\u81ea\u4f53\u306b\u305d\u308c\u307b\u3069\u6642\u9593\u3092\u639b\u3051\u308b\u610f\u5473\u306f\u306a\u3055\u305d\u3046\u3060\u3057\u3002<\/p>\n
\u306a\u306e\u3067\u3001\u6c7a\u5b9a\u6728\u3042\u305f\u308a\u304b\u30af\u30e9\u30b9\u30bf\u30ea\u30f3\u30b0\u3092\u4f7f\u3063\u3066\u3001\u99d2\u306e\u30c4\u30ea\u30fc\uff08\u4eca\u56de\u306f\uff18\u7a2e\u985e\u7a0b\u5ea6\u3060\u3051\u3069\uff09\u3092\u4f5c\u308b\u306e\u304c\u3088\u3044\u306e\u3067\u3057\u3087\u3046\u3002<\/p>\n
<\/a><\/p>\n
\u306e\u3088\u3046\u306b\u3001\u8272\u5473\u3067\u5206\u5272\u3067\u5927\u96d1\u628a\u306b\u5206\u5272\u304c\u3067\u304d\u308b\u306e\u306f\u660e\u3089\u304b\u306a\u306e\u3067\u3001\u306a\u3093\u3089\u304b\u306e\u7279\u5fb4\u91cf\u3092\u5143\u306b\u80cc\u666f\u753b\u50cf\u304b\u3089\u629c\u304d\u51fa\u3057\u305f\u5f8c\u306b\u3001\u76f8\u4e92\u306b\u99d2\u3092\u5224\u65ad\u3059\u308b\u305f\u3081\u306b\u518d\u3073\u7279\u5fb4\u91cf\u3092\u4f7f\u3046\u3068\u3044\u30462\u6bb5\u968e\u306b\u306a\u308b\u306e\u3067\u3057\u3087\u3046\u3002<\/p>\n
\u305d\u3093\u306a\u8a33\u3067\u30aa\u30e9\u30a4\u30ea\u30fc\u306e OpenCV \u3092\u518d\u8aad\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"
\u30c6\u30f3\u30d7\u30ec\u30fc\u30c8\u30de\u30c3\u30c1\u30f3\u30b0\u3067\u753b\u50cf\u304b\u3089\u53d6\u308a\u51fa\u3059\u306b\u306f\u7121\u7406\u304c\u3042\u308a\u305d\u3046\u306a\u306e\u3067\u3001\u3084\u306f\u308a\u3001\u72ec\u81ea\u306a\u30aa\u30d6\u30b8\u30a7\u30af\u30c8\u691c\u51fa\u5668\u3092\u4f5c\u3063\u3066\u307f\u306a\u3044\u3068\u3060\u3081\u304b\u3001\u3068\u601d\u3044\u3001 OpenCV\u3067\u5b66\u3076\u753b\u50cf\u8a8d\u8b58\uff1a\u7b2c4\u56de\u3000\u30aa\u30d6\u30b8\u30a7\u30af\u30c8\u691c\u51fa\u5668\u306e\u4f5c\u6210\u65b9\u6cd5\uff5cgihyo.jp \u2026  … \u7d9a\u304d\u3092\u8aad\u3080 →<\/span><\/a><\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[3,35],"tags":[],"class_list":["post-2537","post","type-post","status-publish","format-standard","hentry","category-dev","category-opencv"],"jetpack_featured_media_url":"","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"http:\/\/www.moonmile.net\/blog\/wp-json\/wp\/v2\/posts\/2537","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/www.moonmile.net\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/www.moonmile.net\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/www.moonmile.net\/blog\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"http:\/\/www.moonmile.net\/blog\/wp-json\/wp\/v2\/comments?post=2537"}],"version-history":[{"count":4,"href":"http:\/\/www.moonmile.net\/blog\/wp-json\/wp\/v2\/posts\/2537\/revisions"}],"predecessor-version":[{"id":2542,"href":"http:\/\/www.moonmile.net\/blog\/wp-json\/wp\/v2\/posts\/2537\/revisions\/2542"}],"wp:attachment":[{"href":"http:\/\/www.moonmile.net\/blog\/wp-json\/wp\/v2\/media?parent=2537"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.moonmile.net\/blog\/wp-json\/wp\/v2\/categories?post=2537"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.moonmile.net\/blog\/wp-json\/wp\/v2\/tags?post=2537"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}