Merged the trunk r8907, r8908

doc/tutorials/introduction/android_binary_package/android_binary_package_using_with_NDK.rst

@@ -158,40 +158,19 @@ then paste the CDT 8.0 repository URL http://download.eclipse.org/tools/cdt/rele
 .. image:: images/eclipse_inst_cdt_2.png
   :alt: Configure builders
   :align: center
-.. important:: This instructions should be applied for each Android project in *Eclipse* workspace having native (JNI) part to build.
 
-#. Right click on your project in :guilabel:`Package Explorer`  window and select :guilabel:`New` -> :guilabel:`Other`:
   
-     .. image:: images/eclipse_cdt_cfg1.png
-        :alt: Configure CDT
-        :align: center
+.. important:: OpenCV 2.4.2 for Android package contains samples projects pre-configured to use CDT Builder for JNI part build via ``ndk-build``.
 
-#. Select :guilabel:`C/C++` -> :guilabel:`Convert to C/C++ project`:
+#. Define the ``NDKROOT`` environment variable containing the path to Android NDK in your system (e.g. **"X:\\Apps\\android-ndk-r8"** or **"/opt/android-ndk-r8"**)
 
-     .. image:: images/eclipse_cdt_cfg2.png
-        :alt: Configure CDT
-        :align: center
+#. CDT Builder is configured for Windows hosts, on Linux or MacOS open `Project Properties` of the projects having JNI part (`face-detection`, `Tutorial 3` and `Tutorial 4`), select :guilabel:`C/C++ Build`   in the left pane, remove **".cmd"** and leave ``"${NDKROOT}/ndk-build"`` in the :guilabel:`Build command`   edit box and click :guilabel:`OK`.
 
-#. Select :guilabel:`"C++"`, :guilabel:`"Makefile Project"`, :guilabel:`"Other Toolchain"`:
-
-     .. image:: images/eclipse_cdt_cfg3.png
-        :alt: Configure CDT
-        :align: center
-
-#. Right click on your project in :guilabel:`Package Explorer`  window and select :guilabel:`Properties`, then :guilabel:`C/C++ Build`  in the left pane.
-     Unckeck :guilabel:`Use default build command` and put ``ndk-build`` invocation in the :guilabel:`Build command` edit box and click :guilabel:`Apply` :
 
      .. image:: images/eclipse_cdt_cfg4.png
         :alt: Configure CDT
         :align: center
 
-#. Select :guilabel:`Builders` in the left pane, select :guilabel:`"CDT Builder"`, press :guilabel:`Edit`  button on the righ and set check-boxes as on the picture below for automatic rebuild of JNI part:
-
-     .. image:: images/eclipse_cdt_cfg5.png
-        :alt: Configure CDT
-        :align: center
-
-
 #. Use menu :guilabel:`Project` -> :guilabel:`Clean...`  to make sure that NDK build is invoked on the project build:
 
     .. image:: images/eclipse_ndk_build.png
@@ -243,7 +222,7 @@ OpenCV binary package includes 3 samples having JNI resources:
 
   This sample illustrates usage of both simple OpenCV face detector via Java API and advanced detection based face tracker via JNI and C++.
   
-Before OpenCV 2.4.2 for Android these projects are not configured to use CDT for building their native part , so you can do it yourself.
+.. important:: Before OpenCV **2.4.2** for Android these projects were not configured to use CDT for building their native part , so you can do it yourself.
 
 Practice: Create an Android application, which uses OpenCV
 ==========================================================

samples/python2/digits.py

@@ -3,8 +3,19 @@ SVN and KNearest digit recognition.
 
 Sample loads a dataset of handwritten digits from 'digits.png'.
 Then it trains a SVN and KNearest classifiers on it and evaluates
-their accuracy. Moment-based image deskew is used to improve 
-the recognition accuracy.
+their accuracy. 
+
+Following preprocessing is applied to the dataset:
+ - Moment-based image deskew (see deskew())
+ - Digit images are split into 4 10x10 cells and 16-bin
+   histogram of oriented gradients is computed for each
+   cell
+ - Transform histograms to space with Hellinger metric (see [1] (RootSIFT))
+
+
+[1] R. Arandjelovic, A. Zisserman
+    "Three things everyone should know to improve object retrieval"
+    http://www.robots.ox.ac.uk/~vgg/publications/2012/Arandjelovic12/arandjelovic12.pdf
 
 Usage:
    digits.py
@@ -14,17 +25,25 @@ import numpy as np
 import cv2
 from multiprocessing.pool import ThreadPool
 from common import clock, mosaic
+from numpy.linalg import norm
 
 SZ = 20 # size of each digit is SZ x SZ
 CLASS_N = 10
 DIGITS_FN = 'digits.png'
 
+def split2d(img, cell_size, flatten=True):
+    h, w = img.shape[:2]
+    sx, sy = cell_size
+    cells = [np.hsplit(row, w//sx) for row in np.vsplit(img, h//sy)]
+    cells = np.array(cells)
+    if flatten:
+        cells = cells.reshape(-1, sy, sx)
+    return cells
+
 def load_digits(fn):
     print 'loading "%s" ...' % fn
     digits_img = cv2.imread(fn, 0)
-    h, w = digits_img.shape
-    digits = [np.hsplit(row, w/SZ) for row in np.vsplit(digits_img, h/SZ)]
-    digits = np.array(digits).reshape(-1, SZ, SZ)
+    digits = split2d(digits_img, (SZ, SZ))
     labels = np.repeat(np.arange(CLASS_N), len(digits)/CLASS_N)
     return digits, labels
 
@@ -92,6 +111,31 @@ def evaluate_model(model, digits, samples, labels):
         vis.append(img)
     return mosaic(25, vis)
 
+def preprocess_simple(digits):
+    return np.float32(digits).reshape(-1, SZ*SZ) / 255.0
+
+def preprocess_hog(digits):
+    samples = []
+    for img in digits:
+        gx = cv2.Sobel(img, cv2.CV_32F, 1, 0)
+        gy = cv2.Sobel(img, cv2.CV_32F, 0, 1)
+        mag, ang = cv2.cartToPolar(gx, gy)
+        bin_n = 16
+        bin = np.int32(bin_n*ang/(2*np.pi))
+        bin_cells = bin[:10,:10], bin[10:,:10], bin[:10,10:], bin[10:,10:]
+        mag_cells = mag[:10,:10], mag[10:,:10], mag[:10,10:], mag[10:,10:]
+        hists = [np.bincount(b.ravel(), m.ravel(), bin_n) for b, m in zip(bin_cells, mag_cells)]
+        hist = np.hstack(hists)
+
+        # transform to Hellinger kernel
+        eps = 1e-7 
+        hist /= hist.sum() + eps
+        hist = np.sqrt(hist)
+        hist /= norm(hist) + eps
+
+        samples.append(hist)
+    return np.float32(samples)
+
 
 if __name__ == '__main__':
     print __doc__
@@ -100,12 +144,12 @@ if __name__ == '__main__':
     
     print 'preprocessing...'
     # shuffle digits
-    rand = np.random.RandomState(12345)
+    rand = np.random.RandomState(321)
     shuffle = rand.permutation(len(digits))
     digits, labels = digits[shuffle], labels[shuffle]
     
     digits2 = map(deskew, digits)
-    samples = np.float32(digits2).reshape(-1, SZ*SZ) / 255.0
+    samples = preprocess_hog(digits2)
     
     train_n = int(0.9*len(samples))
     cv2.imshow('test set', mosaic(25, digits[train_n:]))
@@ -115,13 +159,13 @@ if __name__ == '__main__':
 
     
     print 'training KNearest...'
-    model = KNearest(k=1)
+    model = KNearest(k=4)
     model.train(samples_train, labels_train)
     vis = evaluate_model(model, digits_test, samples_test, labels_test)
     cv2.imshow('KNearest test', vis)
 
     print 'training SVM...'
-    model = SVM(C=4.66, gamma=0.08)
+    model = SVM(C=2.67, gamma=5.383)
     model.train(samples_train, labels_train)
     vis = evaluate_model(model, digits_test, samples_test, labels_test)
     cv2.imshow('SVM test', vis)

samples/python2/digits_adjust.py

@@ -76,7 +76,7 @@ class App(object):
         shuffle = np.random.permutation(len(digits))
         digits, labels = digits[shuffle], labels[shuffle]
         digits2 = map(deskew, digits)
-        samples = np.float32(digits2).reshape(-1, SZ*SZ) / 255.0
+        samples = preprocess_hog(digits2)
         return samples, labels
 
     def get_dataset(self):
@@ -95,8 +95,8 @@ class App(object):
         return ires
             
     def adjust_SVM(self):
-        Cs = np.logspace(0, 5, 10, base=2)
-        gammas = np.logspace(-7, -2, 10, base=2)
+        Cs = np.logspace(0, 10, 15, base=2)
+        gammas = np.logspace(-7, 4, 15, base=2)
         scores = np.zeros((len(Cs), len(gammas)))
         scores[:] = np.nan
 
@@ -114,6 +114,9 @@ class App(object):
             print '%d / %d (best error: %.2f %%, last: %.2f %%)' % (count+1, scores.size, np.nanmin(scores)*100, score*100)
         print scores
 
+        print 'writing score table to "svm_scores.npz"'
+        np.savez('svm_scores.npz', scores=scores, Cs=Cs, gammas=gammas)
+
         i, j = np.unravel_index(scores.argmin(), scores.shape)
         best_params = dict(C = Cs[i], gamma=gammas[j])
         print 'best params:', best_params
@@ -142,7 +145,6 @@ if __name__ == '__main__':
     
     print __doc__
 
-
     args, _ = getopt.getopt(sys.argv[1:], '', ['model=', 'cloud', 'env='])
     args = dict(args)
     args.setdefault('--model', 'svm')

samples/python2/digits_video.py

 import numpy as np
 import cv2
-import digits
 import os
 import video
 from common import mosaic
 
+from digits import *
 
 
 def main():
@@ -15,11 +15,9 @@ def main():
         print '"%s" not found, run digits.py first' % classifier_fn
         return 
     
-    model = digits.SVM()
+    model = SVM()
     model.load('digits_svm.dat')
 
-    SZ = 20
-
     while True:
         ret, frame = cap.read()
         gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
@@ -55,13 +53,12 @@ def main():
             A[:,:2] = np.eye(2)*s
             A[:,2] = t
             sub1 = cv2.warpAffine(sub, A, (SZ, SZ), flags=cv2.WARP_INVERSE_MAP | cv2.INTER_LINEAR)
-            sub1 = digits.deskew(sub1)
+            sub1 = deskew(sub1)
             if x+w+SZ < frame.shape[1] and y+SZ < frame.shape[0]:
                 frame[y:,x+w:][:SZ, :SZ] = sub1[...,np.newaxis]
                 
-            sample = np.float32(sub1).reshape(1,SZ*SZ) / 255.0
+            sample = preprocess_hog([sub1])
             digit = model.predict(sample)[0]
-
             cv2.putText(frame, '%d'%digit, (x, y), cv2.FONT_HERSHEY_PLAIN, 1.0, (200, 0, 0), thickness = 1)
 
 

samples/python2/fitline.py

@@ -2,14 +2,16 @@
 Robust line fitting.
 ==================
 
-Example of using cv2.fitLine function for fitting line to points in presence of outliers.
+Example of using cv2.fitLine function for fitting line 
+to points in presence of outliers.
 
 Usage
 -----
 fitline.py
 
-Switch through different M-estimator functions and see, how well the robust functions
-fit the line even in case of ~50% of outliers.
+Switch through different M-estimator functions and see, 
+how well the robust functions fit the line even 
+in case of ~50% of outliers.
 
 Keys
 ----