def svmEvaluate(model, digits, samples, labels):
predictions = svmPredict(model, samples)
print predictions
accuracy = (labels == predictions).mean()
print('Percentage Accuracy: %.2f %%' % (accuracy * 100))
confusion = np.zeros((CLASS_N + 1, CLASS_N + 1), np.int32)
print predictions.shape
print labels.shape
for i, j in zip(labels, predictions):
confusion[int(i), int(j)] += 1
vis = []
# i = 0
for img, flag in zip(digits, predictions == labels):
img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
if not flag:
img[..., :2] = 0
# print str(unichr(int(labels[i]) + 96))
# print str(unichr(int(predictions[i]) + 96))
# cv2.imshow('dupaia', img)
# cv2.waitKey(0)
# i += 1
vis.append(img)
return mosaic(25, vis)
def generate_svm():
digits, labels = load_digits(DIGITS_FN)
print('preprocessing...')
# shuffle digits
rand = np.random.RandomState(321)
shuffle = rand.permutation(len(digits))
digits, labels = digits[shuffle], labels[shuffle]
digits2 = list(map(deskew, digits))
samples = preprocess_hog(digits2)
train_n = int(0.9*len(samples))
cv2.imshow('test set', mosaic(25, digits[train_n:]))
digits_train, digits_test = np.split(digits2, [train_n])
samples_train, samples_test = np.split(samples, [train_n])
labels_train, labels_test = np.split(labels, [train_n])
print('training SVM...')
model = SVM(C=2.67, gamma=5.383)
model.train(samples_train, labels_train)
vis = evaluate_model(model, digits_test, samples_test, labels_test)
print('saving SVM as "digits_svm.dat"...')
return model
cv2.waitKey(0)
def generate_mosaic(directory):
digits = np.empty([0,SZ,SZ])
prefix = str(SZ)+'/cropped_img'
suffix = '.png'
for i in range(1,11):
for j in range(20,100):
infix= '%03d-%05d' % (i, j)
fname = directory+prefix+infix+suffix
img = cv2.imread(fname, 0)
digits = np.append(digits, np.array([img]),0)
cv2.imwrite('font'+str(SZ)+'.png', mosaic(20, digits))
def evaluate_model(model, digits, samples, labels):
resp = model.predict(samples)
err = (labels != resp).mean()
print('error: %.2f %%' % (err * 100))
confusion = np.zeros((10, 10), np.int32)
for i, j in zip(labels, resp):
confusion[i, int(j)] += 1
print('confusion matrix:')
print(confusion)
print()
vis = []
for img, flag in zip(digits, resp == labels):
img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
if not flag:
img[..., :2] = 0
vis.append(img)
return mosaic(25, vis)
def svmEvaluate(model, digits, samples, labels):
predictions = svmPredict(model, samples)
accuracy = (labels == predictions).mean()
print('Percentage Accuracy: %.2f %%' % (accuracy*100))
confusion = np.zeros((10, 10), np.int32)
for i, j in zip(labels, predictions):
confusion[int(i), int(j)] += 1
print('confusion matrix:')
print(confusion)
vis = []
for img, flag in zip(digits, predictions == labels):
img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
if not flag:
img[...,:2] = 0
vis.append(img)
return mosaic(25, vis)
def svmEvaluate(model, digits, samples, labels):
predictions = svmPredict(model, samples)
accuracy = (labels == predictions).mean()
print('Percentage Accuracy: %.2f %%' % (accuracy*100))
confusion = np.zeros((10, 10), np.int32)
for i, j in zip(labels, predictions):
confusion[int(i), int(j)] += 1
print('confusion matrix:')
print(confusion)
vis = []
for img, flag in zip(digits, predictions == labels):
img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
if not flag:
img[...,:2] = 0
vis.append(img)
return mosaic(25, vis)
def evaluate_model(model, digits, samples, labels):
resp = model.predict(samples)
err = (labels != resp).mean()
print 'error: %.2f %%' % (err*100)
confusion = np.zeros((10, 10), np.int32)
for i, j in zip(labels, resp):
confusion[i, j] += 1
print 'confusion matrix:'
print confusion
print
vis = []
for img, flag in zip(digits, resp == labels):
img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
if not flag:
img[...,:2] = 0
vis.append(img)
return mosaic(25, vis)
def evaluate_model(model, digits, samples, labels):
resp = model.predict(samples)
err = (labels != resp).mean()
print('error: %.2f %%' % (err * 100))
# confusion table, where x is ground-truth and y is predicted
confusion = np.zeros((10, 10), np.int32)
for i, j in zip(labels, resp):
confusion[i, int(j)] += 1
print('confusion matrix:')
print(confusion)
print()
# if not correctly predicted, set 0 all channels except Red to make it red
vis = []
for img, flag in zip(digits, resp == labels):
img = np.uint8((img - img.min()) / (img.max() - img.min()) * 255.)
img = cv.cvtColor(img, cv.COLOR_GRAY2BGR)
if not flag:
img[..., :2] = 0
vis.append(img)
return mosaic(25, vis)
def main():
print __doc__
digits, labels = load_digits(DIGITS_FN)
print 'preprocessing...'
# shuffle digits
rand = np.random.RandomState(321)
shuffle = rand.permutation(len(digits))
digits, labels = digits[shuffle], labels[shuffle]
digits2 = map(deskew, digits)
samples = preprocess_hog(digits2)
train_n = int(0.9*len(samples))
cv2.imshow('test set', mosaic(25, digits[train_n:]))
digits_train, digits_test = np.split(digits2, [train_n])
samples_train, samples_test = np.split(samples, [train_n])
labels_train, labels_test = np.split(labels, [train_n])
print 'training KNearest...'
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=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)
print 'saving SVM as "digits_svm.dat"...'
model.save('digits_svm.dat')
print 'training SimpleAI...'
model = simpleai.SimpleAI(class_n = CLASS_N)
model.train(samples_train, labels_train)
vis = evaluate_model(model, digits_test, samples_test, labels_test)
cv2.imshow('SimpleAI test', vis)
cv2.waitKey(0)
def evaluate_model(model, items, samples, labels):
resp = model.predict(samples)
err = (labels != resp).mean()
print 'error: %.2f %%' % (err*100)
confusion = np.zeros((len(shoeCategory), len(shoeCategory)), np.int32)
for i, j in zip(labels, resp):
confusion[i, j] += 1
print 'confusion matrix:'
print confusion
print
vis = []
for img, guess, actual in zip(items, resp, labels):
img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
cv2.putText(img, idCategory[int(guess)], (0,7), cv2.FONT_HERSHEY_SIMPLEX, 0.35, 255)
cv2.putText(img, idCategory[int(actual)], (0,15), cv2.FONT_HERSHEY_SIMPLEX, 0.35, 255)
# cv2.putText(img,"actual=" + idCategory[int(guess)], (0,15), cv2.FONT_HERSHEY_SIMPLEX, 0.25, 255)
if not guess == actual:
img[...,:2] = 0
vis.append(img)
return mosaic(mosaic_SZ, vis), (err*100)
def evaluate_model(model, items, samples, labels):
resp = model.predict(samples)
err = (labels != resp).mean()
print 'error: %.2f %%' % (err * 100)
confusion = np.zeros((len(shoeCategory), len(shoeCategory)), np.int32)
for i, j in zip(labels, resp):
confusion[i, j] += 1
print 'confusion matrix:'
print confusion
print
vis = []
for img, guess, actual in zip(items, resp, labels):
img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
cv2.putText(img, idCategory[int(guess)], (0, 7),
cv2.FONT_HERSHEY_SIMPLEX, 0.35, 255)
cv2.putText(img, idCategory[int(actual)], (0, 15),
cv2.FONT_HERSHEY_SIMPLEX, 0.35, 255)
# cv2.putText(img,"actual=" + idCategory[int(guess)], (0,15), cv2.FONT_HERSHEY_SIMPLEX, 0.25, 255)
if not guess == actual:
img[..., :2] = 0
vis.append(img)
return mosaic(mosaic_SZ, vis), (err * 100)
print __doc__
args, _ = getopt.getopt(sys.argv[1:], '', ['input='])
args = dict(args)
args.setdefault('--input', 'data/info.dat')
print 'reading file %s' % args['--input']
lines = [line.strip() for line in open(args['--input'])]
vis = []
f = open('./positives.txt', 'w')
for line in lines:
rows = line.split()
source_file_name = rows[0]
print line
if len(rows) != 6:
print line
# print "%d:%d, %d:%d" % (int(rows[2]), int(rows[4])+ int(rows[2]), int(rows[3]), int(rows[5])+ int(rows[3]))
img = cv2.imread('/' + source_file_name, cv2.CV_LOAD_IMAGE_GRAYSCALE)
cr = img[int(rows[3]):int(rows[5]) + int(rows[3]),
int(rows[2]):int(rows[4]) + int(rows[2])]
vis.append(cv2.resize(cr, (50, 50)))
fn = './positive_images/' + source_file_name.split('/')[2]
f.write(fn + '\n')
cv2.imwrite(fn, cr)
cv2.imwrite('out/test_data.jpg', mosaic(50, np.array(vis)))
f.close()
if __name__ == '__main__':
print __doc__
digits, labels = load_digits(DIGITS_FN)
print 'preprocessing...'
# shuffle digits
rand = np.random.RandomState(321)
shuffle = rand.permutation(len(digits))
digits, labels = digits[shuffle], labels[shuffle]
digits2 = map(deskew, digits)
samples = preprocess_hog(digits2)
train_n = int(0.9*len(samples))
cv2.imshow('test set', mosaic(25, digits[train_n:]))
digits_train, digits_test = np.split(digits2, [train_n])
samples_train, samples_test = np.split(samples, [train_n])
labels_train, labels_test = np.split(labels, [train_n])
print 'training KNearest...'
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=2.67, gamma=5.383)
model.train(samples_train, labels_train)
vis = evaluate_model(model, digits_test, samples_test, labels_test)
if __name__ == '__main__':
print 'WORK PATH = ' + os.getcwd()
model_path = r'E:\APC\sample_test\model.npy'
vector_stub_path = r'e:\APC\feature\canditate.vec'
vec_stu = stub.read(vector_stub_path)
test_path = r'E:\APC\sample'
test_files = os.listdir(test_path)
for test_file in test_files:
img = plt.imread(os.path.join(test_path, test_file))[45:246,
160:565, :]
img = cv2.resize(img, (256, 128))
# num_features = len(get_hog_feature(img))
model = load_model(model_path,
input_dim=5000,
hidden_dim=200,
out_dim=2)
scale_set = np.arange(0.8, 1.2, 0.1)
rect_init_size = [50, 30]
step_size = [8, 8]
head_p, head_n = detect(img, model, rect_init_size, scale_set,
step_size, vec_stu)
plt.subplot(1, 2, 1)
plt.imshow(mosaic(10, head_p))
plt.title('HEAD')
plt.subplot(1, 2, 2)
plt.imshow(mosaic(20, head_n))
plt.title('NO HEAD')
plt.show()
#digits = np.concatenate((digits, fonts), axis=0)
#labels = np.concatenate((labels, fontlabels), axis=0)
print digits.shape, labels.shape
print 'preprocessing...'
# shuffle digits
rand = np.random.RandomState(321)
shuffle = rand.permutation(len(digits))
digits, labels = digits[shuffle], labels[shuffle]
digits2 = map(deskew, digits)
samples = preprocess_hog(digits2)
train_n = int(0.9*len(samples))
#train_n = int(1.0*len(samples))
cv2.imshow('train set', mosaic(25, digits[:train_n]))
digits_train, digits_test = np.split(digits2, [train_n])
samples_train, samples_test = np.split(samples, [train_n])
labels_train, labels_test = np.split(labels, [train_n])
digits_img_test = cv2.imread('final'+str(SZ)+'.png', 0)
digits_img_test = cv2.bitwise_not(digits_img_test)
digits_test = split2d(digits_img_test, (SZ, SZ))
digits2_test = map(deskew, digits_test)
samples_test = preprocess_hog(digits2_test)
cv2.imshow('test set', mosaic(9, digits_test))
#cv2.imshow('test set', mosaic(25, digits_test))
print 'training SVM...'
model = SVM(C=2.67, gamma=5.383)
image_source = option_dict['image_source']
shoes, samples, labels = preprocessShoes()
print 'creating mosaic...'
train_n = int(0.9 * len(samples))
shoes_train, shoes_test = np.split(shoes, [train_n])
samples_train, samples_test = np.split(samples, [train_n])
labels_train, labels_test = np.split(labels, [train_n])
for label in shoeCategory.values():
images = [i for i, l in zip(shoes_test, labels_test) if (l == label)]
if (len(images) > 0):
cv2.imwrite('out/test_' + idCategory[label] + '_set.jpg',
mosaic(20, images))
for label in shoeCategory.values():
images = [i for i, l in zip(shoes_test, labels_test) if (l == label)]
if (len(images) > 0):
cv2.imwrite('out/test_' + idCategory[label] + '_set.jpg',
mosaic(20, images))
cv2.imwrite('out/test_set.jpg', mosaic(20, shoes_test))
cv2.imwrite('out/train_set.jpg', mosaic(20, shoes_train))
print 'training KNearest...'
model = KNearest(k=4)
model.train(samples_train, labels_train)
vis, knearestError = evaluate_model(model, shoes_test, samples_test,
labels_test)
#!/usr/bin/python3
import cv2
import sys
from common import mosaic, KNearest, SVM, OCR
from pregrandbux import Grandbux
if __name__ == '__main__':
# model = KNearest(k=4)
model = SVM(C=2.67, gamma=5.383)
ABC = 'abcdefghijklmnopqrstuvwxyz'
ocr = OCR(ABC, model)
do_load = len(sys.argv) > 1
if do_load:
# ocr.train('out-min', 'grandbux_svm.dat', 1.0)
ocr.load('grandbux_svm.dat')
samples = ocr.preprocess_hog(Grandbux(sys.argv[1]).segments(20))
print('SOLVE: {}'.format(''.join(ocr.labels(ocr.predict(samples)))))
else:
print('training SVM...')
ocr.train('out-min', 'grandbux_svm.dat')
cv2.imshow('SVM', mosaic(25, ocr.digits))
cv2.waitKey(0)
# get_hog_feature(cv2.imread(os.path.join(test_path, test_files[0]))))
num_features = len(vec_stu)
num_classes = 2
model = load_model(model_path, input_dim=num_features,
hidden_dim=num_hidden_ep, out_dim=num_classes)
head_p = []
head_n = []
for elem in test_files:
if os.path.splitext(elem)[-1] not in EXT_DICT:
continue
data = cv2.imread(os.path.join(test_path, elem))
start = time()
feature = get_icf_feature(data, vec_stu)
prob, res = predict(model, feature)
print "spend time: %f" % (time() - start)
if res == 1:
head_n.append(cv2.resize(data, (32, 32)))
else:
head_p.append(cv2.resize(data, (32, 32)))
plt.subplot(1,2,1)
plt.imshow(mosaic(20, head_p))#, aspect='auto')
plt.title('HEAD')
plt.subplot(1,2,2)
plt.imshow(mosaic(20, head_n))
plt.title('NO HEAD')
plt.show()
if __name__ == "__main__":
print __doc__
digits, labels = load_digits(DIGITS_FN)
print "preprocessing..."
# shuffle digits
rand = np.random.RandomState(321)
shuffle = rand.permutation(len(digits))
digits, labels = digits[shuffle], labels[shuffle]
digits2 = map(deskew, digits)
samples = preprocess_hog(digits2)
train_n = int(0.9 * len(samples))
cv2.imshow("test set", mosaic(25, digits[train_n:]))
digits_train, digits_test = np.split(digits2, [train_n])
samples_train, samples_test = np.split(samples, [train_n])
labels_train, labels_test = np.split(labels, [train_n])
print "training KNearest..."
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=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)
# process each cell - if the cell has a digit (largest contour area exceeds threshold area), then mark cell for training
labels = []
samples = []
for i in range(81):
cell = cells[i]
contours, hierarchy = cv2.findContours(cell.copy(), cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
if len(contours) > 0:
largestContour = contours[np.argmax(map(cv2.contourArea, contours))]
if cv2.contourArea(largestContour) >= DIGIT_MIN_AREA:
cv2.imshow('What digit is this?', cell)
label = cv2.waitKey(0)-48
if (label >= 1) and (label <= 9):
labels.append(label)
samples.append(cell)
elif (label == 27):
break
cv2.imshow('Acquired data', mosaic(25, samples))
print labels
cv2.waitKey(0)
np.save('data/samples_sudoku' + iteration, samples)
np.save('data/labels_sudoku' + iteration, labels)
print 'Data #' + iteration + ' saved'
# exit
cv2.destroyAllWindows()
print __doc__
args, _ = getopt.getopt(sys.argv[1:], "", ["input="])
args = dict(args)
args.setdefault("--input", "data/info.dat")
print "reading file %s" % args["--input"]
lines = [line.strip() for line in open(args["--input"])]
vis = []
f = open("./positives.txt", "w")
for line in lines:
rows = line.split()
source_file_name = rows[0]
print line
if len(rows) != 6:
print line
# print "%d:%d, %d:%d" % (int(rows[2]), int(rows[4])+ int(rows[2]), int(rows[3]), int(rows[5])+ int(rows[3]))
img = cv2.imread("/" + source_file_name, cv2.CV_LOAD_IMAGE_GRAYSCALE)
cr = img[int(rows[3]) : int(rows[5]) + int(rows[3]), int(rows[2]) : int(rows[4]) + int(rows[2])]
vis.append(cv2.resize(cr, (50, 50)))
fn = "./positive_images/" + source_file_name.split("/")[2]
f.write(fn + "\n")
cv2.imwrite(fn, cr)
cv2.imwrite("out/test_data.jpg", mosaic(50, np.array(vis)))
f.close()
print('preprocessing...')
# shuffle digits
rand = np.random.RandomState(321) #随机种子321
shuffle = rand.permutation(len(digits))
digits, labels = digits[shuffle], labels[shuffle] #打乱数字顺序
# deskew
deskewed_digits = list(map(deskew, digits)) #纠正图片倾斜
# hog
samples = preprocess_hog(deskewed_digits) #计算hog特征
# split into validation sets
train_n = int(0.8 * len(samples)) #划分训练测试集
print(train_n)
# what is mosaic?
show('test set', mosaic(25, digits[train_n:]))
# the original digits
digits_train, digits_test = np.split(deskewed_digits, [train_n])
# the descriptors
samples_train, samples_test = np.split(samples, [train_n])
labels_train, labels_test = np.split(labels, [train_n])
print('training KNearest...') #knn分类器
model = KNearest(k=4)
model.train(samples_train, labels_train) # provide x, y
vis = evaluate_model(model, digits_test, samples_test, labels_test)
show('KNearest test', vis)
print('training SVM...') #SVM分类器
model = SVM(C=2.67, gamma=5.383)
databaseUri = option_dict['database']
image_source = option_dict['image_source']
shoes, samples, labels = preprocessShoes()
print 'creating mosaic...'
train_n = int(0.9*len(samples))
shoes_train, shoes_test = np.split(shoes, [train_n])
samples_train, samples_test = np.split(samples, [train_n])
labels_train, labels_test = np.split(labels, [train_n])
for label in shoeCategory.values():
images = [i for i, l in zip(shoes_test, labels_test) if (l == label)]
if (len(images) > 0):
cv2.imwrite('out/test_' + idCategory[label] + '_set.jpg', mosaic(20, images))
for label in shoeCategory.values():
images = [i for i, l in zip(shoes_test, labels_test) if (l == label)]
if (len(images) > 0):
cv2.imwrite('out/test_' + idCategory[label] + '_set.jpg', mosaic(20, images))
cv2.imwrite('out/test_set.jpg', mosaic(20, shoes_test))
cv2.imwrite('out/train_set.jpg', mosaic(20, shoes_train))
print 'training KNearest...'
model = KNearest(k=4)
model.train(samples_train, labels_train)
vis, knearestError = evaluate_model(model, shoes_test, samples_test, labels_test)
cv2.imwrite('out/KNearest_test_' + str(SZ) + '.jpg', vis)
ri = i/10
ci = i%10
if ci != 9 and ri!=9:
src = bm[ri:ri+2, ci:ci+2 , :].reshape((4,2))
dst = np.array( [ [ci*50,ri*50],[(ci+1)*50-1,ri*50],[ci*50,(ri+1)*50-1],[(ci+1)*50-1,(ri+1)*50-1] ], np.float32)
retval = cv2.getPerspectiveTransform(src,dst)
warp = cv2.warpPerspective(res2,retval,(450,450))
output[ri*50:(ri+1)*50-1 , ci*50:(ci+1)*50-1] = warp[ri*50:(ri+1)*50-1 , ci*50:(ci+1)*50-1].copy()
output = cv2.cvtColor(output,cv2.COLOR_BGR2GRAY)
output = cv2.GaussianBlur(output,(5,5),0)
output = cv2.adaptiveThreshold(output,255,cv2.ADAPTIVE_THRESH_MEAN_C,cv2.THRESH_BINARY,3,2)
t = time.clock()-start
imglist = []
print '(%.2f seconds)\n' % t
for i in range(0,450,50):
for j in range(0,450,50):
#cv2.imshow('cell', output[i+2:i+48, j+2:j+48])
temp = output[i+2:i+48, j+2:j+48]
temp = cv2.resize(temp,(20,20), interpolation = cv2.INTER_CUBIC)
imglist.append(temp)
#cv2.waitKey(0)
from common import mosaic
cv2.imwrite('final.png', mosaic(9, imglist))
cv2.imshow('final', mosaic(9, imglist))
cv2.waitKey(0)
cv2.imwrite('output.png', output)
cv2.imshow('img', output)
cv2.waitKey(0)
cv2.destroyAllWindows()
#!/usr/bin/python3
import cv2
import sys
from common import mosaic, KNearest, SVM, OCR
from pregrandbux import Grandbux
if __name__ == "__main__":
# model = KNearest(k=4)
model = SVM(C=2.67, gamma=5.383)
ABC = "abcdefghijklmnopqrstuvwxyz"
ocr = OCR(ABC, model)
do_load = len(sys.argv) > 1
if do_load:
# ocr.train('out-min', 'grandbux_svm.dat', 1.0)
ocr.load("grandbux_svm.dat")
samples = ocr.preprocess_hog(Grandbux(sys.argv[1]).segments(20))
print("SOLVE: {}".format("".join(ocr.labels(ocr.predict(samples)))))
else:
print("training SVM...")
ocr.train("out-min", "grandbux_svm.dat")
cv2.imshow("SVM", mosaic(25, ocr.digits))
cv2.waitKey(0)
labels_train_unrolled = unroll_responses(labels_train, CLASS_N)
model.train(digits_train, labels_train_unrolled, None, params=params)
model.save('dig_nn.dat')
model.load('dig_nn.dat')
def evaluate(model, samples, labels):
'''Evaluates classifier preformance on a given labeled samples set.'''
ret, resp = model.predict(samples)
resp = resp.argmax(-1)
error_mask = (resp == labels)
accuracy = error_mask.mean()
return accuracy, error_mask
# evaluate model
train_accuracy, _ = evaluate(model, digits_train, labels_train)
print 'train accuracy: ', train_accuracy
test_accuracy, test_error_mask = evaluate(model, digits_test, labels_test)
print 'test accuracy: ', test_accuracy
# visualize test results
vis = []
for img, flag in zip(digits_test, test_error_mask):
img = np.uint8(img).reshape(SZ, SZ)
img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
if not flag:
img[...,:2] = 0
vis.append(img)
vis = mosaic(25, vis)
cv2.imshow('test', vis)
cv2.waitKey()
