def __init__(self, patch, id):
Feature.__init__(self, patch, id)
self.HUE_MODEL = np.array([
0., 0.00091991, 0.01559402, 0.06014748, 0.20452348, 0.55954024,
0.15927488, 0., 0., 0., 0., 0., 0., 0., 0., 0.
])
self.HUE_MODEL = normalize(self.HUE_MODEL, norm='l1')[0]
def __init__(self, patch, id):
Feature.__init__(self, patch, id)
self.FEATURE_MODEL = utils.loadArray(
"FeatureHeartShapeContourModel.npz")
self.FEATURE_MODEL_NUM_CONTOURS = 2
self._checkRep()
def __init__(self, patch, id): Feature.__init__(self, patch, id) self.HISTBINNUM = 16 self.FEATURE_MODEL = np.array([ 0., 0., 0., 0., 0.04649839, 0.13987842, 0.09450414, 0., 0., 0., 0., 0., 0., 0., 0., 0. ]) self.FEATURE_MODEL = np.concatenate((self.FEATURE_MODEL, np.zeros(len(range(4,7)))), axis = 1) self.FEATURE_MODEL = normalize(self.FEATURE_MODEL, norm='l1')[0] # normalize the histogram using l1
def __init__(self, patch, id): Feature.__init__(self, patch, id) self.HISTBINNUM = 16 # self.FEATURE_MODEL = np.array([0.00351539, 0.57596057, 0.06858266, 0.02048219, 0.02166264, 0.02258441, # 0.01227705, 0.01083268, 0.00764063, 0.01532651, 0.0044623, 0., 0., # 0., 0., 0., 0., 0., 0. ]) self.FEATURE_MODEL = np.zeros(14 + 3) self.FEATURE_MODEL[0] = 1.0 self.FEATURE_MODEL = normalize(self.FEATURE_MODEL, norm='l1')[0] # normalize the histogram using l1
def __init__(self, patch, id): Feature.__init__(self, patch, id) # other than the cutted bins of wanted HOG direction, all other HOG bins' values are zero self.FEATURE_MODEL_INNER = None self.FEATURE_MODEL_BORDER = None self.inner_HOG = None self.border_HOG = None self.GAUSSIAN_SCALE = 2 self.GAUSSIAN_WINDOW_LENGTH_SIGMA = 4.0 # used for border feature case, window length = 4 sigma
def __init__(self, patch, id):
Feature.__init__(self, patch, id)
self.FEATURE_MODEL = np.array([
0., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0.34, 0.33, 0.33, 0., 0., 0.
])
self.FEATURE_MODEL = np.concatenate(
(self.FEATURE_MODEL, np.zeros(3 * len(range(1, 2)))), axis=1)
self.FEATURE_MODEL = normalize(
self.FEATURE_MODEL,
norm='l1')[0] # normalize the FEATURE_MODEL using l1
def __init__(self, patch, id):
Feature.__init__(self, patch, id)
# other than the cutted bins of wanted HOG direction, all other HOG bins' values are zero
self.FEATURE_MODEL = np.array([
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.05131246, 0.05356285,
0., 0.067983, 0.04214793, 0., 0., 0., 0., 0., 0., 0.06094781, 0.,
0., 0., 0., 0., 0., 0.1160895, 0., 0., 0., 0., 0., 0.
])
self.FEATURE_MODEL = normalize(
self.FEATURE_MODEL,
norm='l1')[0] # normalize the histogram using l1
self._checkRep()
def __init__(self, patch, id):
Feature.__init__(self, patch, id)
self.HISTBINNUM = 16
self.GAUSSIAN_SCALE = 3 # default is 3, will detect from 3 to 1
self.GAUSSIAN_WINDOW_LENGTH_SIGMA = 4.0 # used for border feature case, window length = 4 sigma
self.HUE_START_INDEX = None
self.HUE_END_INDEX = None
self.SATURATION_START_INDEX = None
self.SATURATION_END_INDEX = None
self.SATURATION_FILTER_START_INDEX = None
self.SATURATION_FILTER_END_INDEX = None
self.FEATURE_MODEL = None
def __init__(self, patch, id):
Feature.__init__(self, patch, id)
self.FEATURE_MODEL = np.array([
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0.,
0., 0., 0.19443516, 0.23649953, 0.13200154, 0.23102906, 0.15323535,
0.05279936, 0., 0., 0., 0., 0., 0., 0.
])
# reinforce that the other sub patches must not have the hue/saturation for the target color
self.FEATURE_MODEL = np.concatenate(
(self.FEATURE_MODEL,
np.zeros(3 * (len(range(15, 16)) + len(range(3, 9))))),
axis=1)
self.FEATURE_MODEL = normalize(
self.FEATURE_MODEL,
norm='l1')[0] # normalize the FEATURE_MODEL using l1
def __init__(self, patch, id): Feature.__init__(self, patch, id) # the distribution of hue and saturation needed for the subpatch of interest # self.FEATURE_MODEL = np.array([ 0., 0., 0., 0.08106389, 0.11229446, 0., 0., # 0., 0., 0., 0., 0., 0., 0., # 0., 0., 0., 0., 0., 0., 0., # 0., 0., 0.02104336, 0.0664095, 0.02171191, 0., 0., # 0., 0., 0., 0. ]) # hue model and saturation model normalized seperately: self.FEATURE_MODEL = np.array([ 0., 0., 0., 0.15622393, 0.84377607, 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.0]) # reinforce that the other sub patches must not have the hue/saturation for the green color self.FEATURE_MODEL = np.concatenate((self.FEATURE_MODEL, np.zeros(3 *(len(range(3,5))))), axis = 1) self.FEATURE_MODEL = normalize(self.FEATURE_MODEL, norm='l1')[0] # normalize the FEATURE_MODEL using l1
def __init__(self, patch, id):
Feature.__init__(self, patch, id)
self.HISTBINNUM = 16
self.FEATURE_MODEL = np.array([
0.00000000e+00, 1.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 7.38737277e-04, 4.13136744e-01,
5.86124519e-01, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00
])
self.FEATURE_MODEL = np.concatenate(
(self.FEATURE_MODEL, np.zeros(3 * len(range(1, 2)))), axis=1)
self.FEATURE_MODEL = normalize(
self.FEATURE_MODEL,
norm='l1')[0] # normalize the FEATURE_MODEL using l1
def __init__(self, patch, id):
Feature.__init__(self, patch, id)
self.HISTBINNUM = 16
self.FEATURE_MODEL_HUE = np.array([
0., 0., 0.44326144, 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0.
])
self.FEATURE_MODEL_SATURATION = np.array([
0., 0., 0., 0., 0., 0., 0., 0.21110128, 0.21070874, 0., 0., 0., 0.,
0., 0., 0.
])
self.FEATURE_MODEL = np.concatenate(( \
self.FEATURE_MODEL_HUE, \
self.FEATURE_MODEL_SATURATION, \
np.zeros(len(range(2,3)) + len(range(7,9)))), axis = 1) # append the expected border response
self.FEATURE_MODEL = normalize(
self.FEATURE_MODEL,
norm='l1')[0] # normalize the histogram using l1
def __init__(self, patch, id):
Feature.__init__(self, patch, id)
self.HUE_START_INDEX = 8
self.HUE_END_INDEX = 9
self.SATURATION_START_INDEX = 7
self.SATURATION_END_INDEX = 12
self.OTHER_HUE_START_INDEX = 3
self.OTHER_HUE_END_INDEX = 4
self.OTHER_SATURATION_START_INDEX = 6
self.OTHER_SATURATION_END_INDEX = 10
# bottom left blue: hue model and saturation model normalized seperately:
self.BLUE_FEATURE_MODEL = np.array([ 0., 0., 0., 0., 0., 0. , 0.,
0., 1., 0. , 0., 0., 0., 0.,
0., 0. , \
0. , 0., 0., 0., 0.,
0., 0., 0.2, 0.2, 0.2, 0.2,
0.2, 0., 0., 0., 0. ])
# reinforce that the other sub patches must be green
self.GREEN_FEATURE_MODEL = np.array([ 0., 0., 0., 1., 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0.,
0., 0., \
0., 0., 0., 0., 0., 0., 0.25,
0.25, 0.25, 0.25, 0., 0., 0., 0.,
0., 0. ]) # with this, can get the target patch as 4th best
# """try changing the sum of targeted saturation to be = 1"""
# self.BLUE_FEATURE_MODEL = np.array([ 0., 0., 0., 0., 0., 0. , 0.,
# 0., 1., 0. , 0., 0., 0., 0.,
# 0., 0. , \
# 1. ])
# # reinforce that the other sub patches must be green
# self.GREEN_FEATURE_MODEL = np.array([ 0., 0., 0., 1., 0., 0., 0.,
# 0., 0., 0., 0., 0., 0., 0.,
# 0., 0., \
# 1. ])
self.FEATURE_MODEL = np.concatenate(\
(self.BLUE_FEATURE_MODEL, self.GREEN_FEATURE_MODEL, self.GREEN_FEATURE_MODEL, self.GREEN_FEATURE_MODEL), \
axis = 1)
self.FEATURE_MODEL = normalize(
self.FEATURE_MODEL,
norm='l1')[0] # normalize the FEATURE_MODEL using l1
def __init__(self, patch, id):
Feature.__init__(self, patch, id)
self.HISTBINNUM = 16 # adjustable number of bins
self.GAUSSIAN_SCALE = 3
self.GAUSSIAN_WINDOW_LENGTH_SIGMA = 4.0 # used for border feature case, window length = 4 sigma
self.HUE_START_INDEX = None
self.HUE_END_INDEX = None
self.SATURATION_START_INDEX = None
self.SATURATION_END_INDEX = None
"""
TODO 0: try aggragate Hue bins
1: try seperate indexes used to filter hue bins that are wrongly categorized within target bins (now saturation range is very wide, may introduce noise)
2: try change HISTBINNUM to 16
3: If border saturation is concentrated, then add that as well.
4: !try use Saturation * Hue as FEATURE_MODEL
5: try 2D histogram -> leads to the problem of dissimilarity metric
6: !try switch back to 36 bins, but for hue, widen up for border error detection
7: !! if current one still wrongly match, try adjust border_hist weight and increase border_hist saturation range!!!!
"""
self.SATURATION_FILTER_START_INDEX = None
self.SATURATION_FILTER_END_INDEX = None
self.SATURATION_BORDER_FILTER_START_INDEX = None
self.SATURATION_BORDER_FILTER_END_INDEX = None
self.FEATURE_MODEL_HUE = np.zeros(self.HISTBINNUM)
self.FEATURE_MODEL_SATURATION = np.zeros(self.HISTBINNUM)
self.FEATURE_MODEL = None
self.border_hist_hue = None
self.border_hist_saturation = None
self.inner_hist_hue = None
self.inner_hist_saturation = None
# cutted and mannually crafted hist
self.border_hist = None
def __init__(self, patch, id):
Feature.__init__(self, patch, id)
self.HISTBINNUM = 16
self.GAUSSIAN_SCALE = 3
self.GAUSSIAN_WINDOW_LENGTH_SIGMA = 6.0 # window length = 6 sigma
self.HUE_START_INDEX = None
self.HUE_END_INDEX = None
self.SATURATION_START_INDEX = None
self.SATURATION_END_INDEX = None
"""TODO: try indexes used to filter hue bins that are wrongly categorized within target bins"""
self.SATURATION_FILTER_START_INDEX = None
self.SATURATION_FILTER_END_INDEX = None
### The patch of interest, one of TOP_LEFT, TOP_RIGHT, BOTTOM_LEFT, BOTTOM_RIGHT ###
self.SUBPATCH_OF_INTEREST_INDEX = None
self.FEATURE_MODEL_HUE = np.zeros(self.HISTBINNUM)
self.FEATURE_MODEL_SATURATION = np.zeros(self.HISTBINNUM)
self.FEATURE_MODEL = None
def __init__(self, patch, id): Feature.__init__(self, patch, id) self.HISTBINNUM = 16 self.HUE_START_INDEX = 9 self.HUE_END_INDEX = 10 self.SATURATION_START_INDEX = 8 self.SATURATION_END_INDEX = 10 self.FEATURE_MODEL_HUE = np.zeros(self.HISTBINNUM) self.FEATURE_MODEL_HUE[self.HUE_START_INDEX] = 1.0 # all in bin 9 self.FEATURE_MODEL_SATURATION = np.zeros(self.HISTBINNUM) self.FEATURE_MODEL_SATURATION[self.SATURATION_START_INDEX] = 0.5 # all in bin 8 or bin 9 self.FEATURE_MODEL_SATURATION[self.SATURATION_START_INDEX + 1] = 0.5 # all in bin 8 or bin 9 self.FEATURE_MODEL = np.concatenate(( \ self.FEATURE_MODEL_HUE, \ self.FEATURE_MODEL_SATURATION, \ np.zeros(len(range(self.HUE_START_INDEX,self.HUE_END_INDEX)) + \ len(range(self.SATURATION_START_INDEX,self.SATURATION_END_INDEX)))), axis = 1) # append the expected border response self.FEATURE_MODEL = normalize(self.FEATURE_MODEL, norm='l1')[0] # normalize the histogram using l1
def __init__(self, patch, id):
Feature.__init__(self, patch, id)
self.max_response = None
self._checkRep()
