質問編集履歴
2
変更部分の抽出
test
CHANGED
|
File without changes
|
test
CHANGED
|
@@ -370,8 +370,24 @@
|
|
|
370
370
|
|
|
371
371
|
# 追記1
|
|
372
372
|
|
|
373
|
+
ベクトルの1次元変換は、このような感じでは無いでしょうか？
|
|
374
|
+
|
|
373
375
|
```python
|
|
374
376
|
|
|
377
|
+
# 次元を１次元に変換
|
|
378
|
+
|
|
379
|
+
img_np = np.asarray(img)
|
|
380
|
+
|
|
381
|
+
img_np = img_np.shape[0] * img_np.shape[1]
|
|
382
|
+
|
|
383
|
+
```
|
|
384
|
+
|
|
385
|
+
|
|
386
|
+
|
|
387
|
+
以下、全文
|
|
388
|
+
|
|
389
|
+
```python
|
|
390
|
+
|
|
375
391
|
#!/usr/bin/env python
|
|
376
392
|
|
|
377
393
|
# -*- coding: utf-8 -*-
|
1
回答に基づくデータの追記
test
CHANGED
|
File without changes
|
test
CHANGED
|
@@ -363,3 +363,315 @@
|
|
|
363
363
|
- MacOSX Yosemite 10.10.5
|
|
364
364
|
|
|
365
365
|
- Python 2.7.10
|
|
366
|
+
|
|
367
|
+
|
|
368
|
+
|
|
369
|
+
|
|
370
|
+
|
|
371
|
+
# 追記1
|
|
372
|
+
|
|
373
|
+
```python
|
|
374
|
+
|
|
375
|
+
#!/usr/bin/env python
|
|
376
|
+
|
|
377
|
+
# -*- coding: utf-8 -*-
|
|
378
|
+
|
|
379
|
+
|
|
380
|
+
|
|
381
|
+
import sys
|
|
382
|
+
|
|
383
|
+
import os
|
|
384
|
+
|
|
385
|
+
from PIL import Image
|
|
386
|
+
|
|
387
|
+
import ImageFilter
|
|
388
|
+
|
|
389
|
+
import numpy as np
|
|
390
|
+
|
|
391
|
+
from sklearn import svm
|
|
392
|
+
|
|
393
|
+
|
|
394
|
+
|
|
395
|
+
|
|
396
|
+
|
|
397
|
+
# 定数定義場
|
|
398
|
+
|
|
399
|
+
STANDARD_SIZE = (300, 167)
|
|
400
|
+
|
|
401
|
+
CONST_NUM = 50100
|
|
402
|
+
|
|
403
|
+
|
|
404
|
+
|
|
405
|
+
|
|
406
|
+
|
|
407
|
+
# 変数定義場
|
|
408
|
+
|
|
409
|
+
Images = [] # 教師画像
|
|
410
|
+
|
|
411
|
+
Labels = [] # ラベル（ok = 0, ng = 1とする）
|
|
412
|
+
|
|
413
|
+
|
|
414
|
+
|
|
415
|
+
|
|
416
|
+
|
|
417
|
+
# 学習用画像データ（OK）を読み込む
|
|
418
|
+
|
|
419
|
+
for f in os.listdir("./ok/"):
|
|
420
|
+
|
|
421
|
+
# 画像を読み込む
|
|
422
|
+
|
|
423
|
+
img = Image.open("./ok/" + f)
|
|
424
|
+
|
|
425
|
+
|
|
426
|
+
|
|
427
|
+
# グレースケール化
|
|
428
|
+
|
|
429
|
+
img = img.convert('L')
|
|
430
|
+
|
|
431
|
+
|
|
432
|
+
|
|
433
|
+
# エッジ抽出
|
|
434
|
+
|
|
435
|
+
img = img.filter(ImageFilter.FIND_EDGES)
|
|
436
|
+
|
|
437
|
+
|
|
438
|
+
|
|
439
|
+
# ダウンサイジング
|
|
440
|
+
|
|
441
|
+
img = img.resize(STANDARD_SIZE)
|
|
442
|
+
|
|
443
|
+
|
|
444
|
+
|
|
445
|
+
# 次元を１次元に変換
|
|
446
|
+
|
|
447
|
+
img_np = np.asarray(img)
|
|
448
|
+
|
|
449
|
+
img_np = img_np.shape[0] * img_np.shape[1]
|
|
450
|
+
|
|
451
|
+
|
|
452
|
+
|
|
453
|
+
# データを追加
|
|
454
|
+
|
|
455
|
+
Images.append(img_np)
|
|
456
|
+
|
|
457
|
+
|
|
458
|
+
|
|
459
|
+
# ラベルを設定
|
|
460
|
+
|
|
461
|
+
Labels.append(0)
|
|
462
|
+
|
|
463
|
+
|
|
464
|
+
|
|
465
|
+
print "Images, Labels = " + str(len(Images)) + ", " + str(len(Labels))
|
|
466
|
+
|
|
467
|
+
|
|
468
|
+
|
|
469
|
+
img.save('./out/ok_' + f)
|
|
470
|
+
|
|
471
|
+
|
|
472
|
+
|
|
473
|
+
|
|
474
|
+
|
|
475
|
+
|
|
476
|
+
|
|
477
|
+
# 学習用画像データ（NG）を読み込む
|
|
478
|
+
|
|
479
|
+
for f in os.listdir("./ng/"):
|
|
480
|
+
|
|
481
|
+
# 画像を読み込む
|
|
482
|
+
|
|
483
|
+
img = Image.open("./ng/" + f)
|
|
484
|
+
|
|
485
|
+
|
|
486
|
+
|
|
487
|
+
# グレースケール化
|
|
488
|
+
|
|
489
|
+
img = img.convert('L')
|
|
490
|
+
|
|
491
|
+
|
|
492
|
+
|
|
493
|
+
# エッジ抽出
|
|
494
|
+
|
|
495
|
+
img = img.filter(ImageFilter.FIND_EDGES)
|
|
496
|
+
|
|
497
|
+
|
|
498
|
+
|
|
499
|
+
# ダウンサイジング
|
|
500
|
+
|
|
501
|
+
img = img.resize(STANDARD_SIZE)
|
|
502
|
+
|
|
503
|
+
|
|
504
|
+
|
|
505
|
+
# 次元を１次元に変換
|
|
506
|
+
|
|
507
|
+
img_np = np.asarray(img)
|
|
508
|
+
|
|
509
|
+
img_np = img_np.shape[0] * img_np.shape[1]
|
|
510
|
+
|
|
511
|
+
|
|
512
|
+
|
|
513
|
+
# データを追加
|
|
514
|
+
|
|
515
|
+
Images.append(img_np)
|
|
516
|
+
|
|
517
|
+
|
|
518
|
+
|
|
519
|
+
img.save('./out/ng_' + f)
|
|
520
|
+
|
|
521
|
+
|
|
522
|
+
|
|
523
|
+
print "Images, Labels = " + str(len(Images)) + ", " + str(len(Labels))
|
|
524
|
+
|
|
525
|
+
|
|
526
|
+
|
|
527
|
+
# ラベルを設定
|
|
528
|
+
|
|
529
|
+
Labels.append(1)
|
|
530
|
+
|
|
531
|
+
|
|
532
|
+
|
|
533
|
+
|
|
534
|
+
|
|
535
|
+
|
|
536
|
+
|
|
537
|
+
# ガウシアンカーネルによる SVM インスタンス生成
|
|
538
|
+
|
|
539
|
+
clf = svm.SVC(C=100, kernel="rbf")
|
|
540
|
+
|
|
541
|
+
|
|
542
|
+
|
|
543
|
+
# 教師データで学習 (データに対するフィッティング)
|
|
544
|
+
|
|
545
|
+
print "----------------"
|
|
546
|
+
|
|
547
|
+
print "TRAINING NOW..."
|
|
548
|
+
|
|
549
|
+
print "----------------"
|
|
550
|
+
|
|
551
|
+
|
|
552
|
+
|
|
553
|
+
print "画像数：" + str(len(Images))
|
|
554
|
+
|
|
555
|
+
print "ラベル数：" + str(len(Labels))
|
|
556
|
+
|
|
557
|
+
|
|
558
|
+
|
|
559
|
+
# トレーニング
|
|
560
|
+
|
|
561
|
+
clf.fit(Images, Labels)
|
|
562
|
+
|
|
563
|
+
|
|
564
|
+
|
|
565
|
+
|
|
566
|
+
|
|
567
|
+
# 分類する画像データを入れる変数
|
|
568
|
+
|
|
569
|
+
studentImages = []
|
|
570
|
+
|
|
571
|
+
|
|
572
|
+
|
|
573
|
+
# 分類する画像を読み込む
|
|
574
|
+
|
|
575
|
+
for f in os.listdir("./img/"):
|
|
576
|
+
|
|
577
|
+
print "filename: " + f
|
|
578
|
+
|
|
579
|
+
|
|
580
|
+
|
|
581
|
+
# 画像を読み込む
|
|
582
|
+
|
|
583
|
+
img = Image.open("./img/" + f)
|
|
584
|
+
|
|
585
|
+
|
|
586
|
+
|
|
587
|
+
# グレースケール化
|
|
588
|
+
|
|
589
|
+
img = img.convert('L')
|
|
590
|
+
|
|
591
|
+
|
|
592
|
+
|
|
593
|
+
# エッジ抽出
|
|
594
|
+
|
|
595
|
+
img = img.filter(ImageFilter.FIND_EDGES)
|
|
596
|
+
|
|
597
|
+
|
|
598
|
+
|
|
599
|
+
# ダウンサイジング
|
|
600
|
+
|
|
601
|
+
img = img.resize(STANDARD_SIZE)
|
|
602
|
+
|
|
603
|
+
|
|
604
|
+
|
|
605
|
+
# 次元を１次元に変換
|
|
606
|
+
|
|
607
|
+
img_np = np.asarray(img)
|
|
608
|
+
|
|
609
|
+
img_np = img_np.shape[0] * img_np.shape[1]
|
|
610
|
+
|
|
611
|
+
|
|
612
|
+
|
|
613
|
+
# データを追加
|
|
614
|
+
|
|
615
|
+
studentImages.append(img_np)
|
|
616
|
+
|
|
617
|
+
|
|
618
|
+
|
|
619
|
+
img.save('./out/student_' + f)
|
|
620
|
+
|
|
621
|
+
|
|
622
|
+
|
|
623
|
+
|
|
624
|
+
|
|
625
|
+
print "-------------------"
|
|
626
|
+
|
|
627
|
+
print "CLUSTERING NOW..."
|
|
628
|
+
|
|
629
|
+
print "-------------------"
|
|
630
|
+
|
|
631
|
+
|
|
632
|
+
|
|
633
|
+
# 画像データを分類する
|
|
634
|
+
|
|
635
|
+
results = clf.predict(studentImages)
|
|
636
|
+
|
|
637
|
+
|
|
638
|
+
|
|
639
|
+
for r in results:
|
|
640
|
+
|
|
641
|
+
if r == 0:
|
|
642
|
+
|
|
643
|
+
print "ok"
|
|
644
|
+
|
|
645
|
+
elif r == 1:
|
|
646
|
+
|
|
647
|
+
print "ng"
|
|
648
|
+
|
|
649
|
+
else:
|
|
650
|
+
|
|
651
|
+
print "error"
|
|
652
|
+
|
|
653
|
+
|
|
654
|
+
|
|
655
|
+
```
|
|
656
|
+
|
|
657
|
+
|
|
658
|
+
|
|
659
|
+
|
|
660
|
+
|
|
661
|
+
# 追記2
|
|
662
|
+
|
|
663
|
+
```
|
|
664
|
+
|
|
665
|
+
/Library/Python/2.7/site-packages/sklearn/utils/validation.py:386: DeprecationWarning:
|
|
666
|
+
|
|
667
|
+
Passing 1d arrays as data is deprecated in 0.17 and willraise ValueError in 0.19.
|
|
668
|
+
|
|
669
|
+
Reshape your data either using X.reshape(-1, 1) if your data has a single feature or X.reshape(1, -1) if it contains a single sample.
|
|
670
|
+
|
|
671
|
+
|
|
672
|
+
|
|
673
|
+
ValueError: X and y have incompatible shapes.
|
|
674
|
+
|
|
675
|
+
X has 1 samples, but y has 170.
|
|
676
|
+
|
|
677
|
+
```
|