def setTFList(imageList, centroids, K):
"""
This method used to set tf list to each image. We use centroids and compare
distance between sift descriptors in image, and assign these features to
visual words.
[image, image,..], [[int, int,...],...] => [image, image,..]
"""
processGap = len(imageList)/10.0
processIndex = 0
imageCount = 0
print 'PROCESS Images ',
for image in imageList:
if imageCount == len(imageList) - 1:
print '#'
elif imageCount > processIndex*processGap:
sys.stdout.write('#')
processIndex = processIndex + 1
imageCount = imageCount + 1
siftDescriptor = image.SIFTdescr
wordList = []
# get distances between SIFT and centers
distances = getDistances(centroids, siftDescriptor)
for siftIndex in xrange(len(siftDescriptor)):
minIndex = np.argmin(distances[siftIndex])
wordList.append(minIndex)
# save word list into test image
image.setWordList(wordList)
# get tf list
tfList = getTFlist(wordList, K)
image.settfList(tfList)
return imageList
def setTFListTrainImageFast(trainImageList, labels, K):
"""
This method used to use K-means labels result to set tf list to train images.
[image, image,...], [[int, int,...]] => [image, image,..]
"""
start = 0
for trainImage in trainImageList:
descr = trainImage.SIFTdescr
trainImage.setWordList(labels[start : start + len(descr)])
wordList = trainImage.wordList
tfList = getTFlist(wordList, K)
trainImage.settfList(tfList)
start = start + len(descr)
return trainImageList
