def compareAll(self, rawdatas, parts=200, chunksize=100, threshold=1.0, maxStop=None, jumpsize=10, jumpsize2=100):
if maxStop == None:
maxStop = len(rawdatas)*500
datalist = []
for rawdata in rawdatas:
datasize = int((int(len(rawdata)/parts))/2)*2
datalist.append(rawdata[0:datasize])
for i in range(1, parts):
datalist.append(rawdata[(datasize*(i)):(datasize*(i+1))])
rebuiltdata = b''
simMat = np.zeros((len(datalist),len(datalist)))
print ("CALCULATING")
bar = progress_bar(len(datalist))
for i in range(len(datalist)):
bar.update_bar(i)
for j in range(len(datalist)):
if i > j:
simMat[i][j] = 0
continue
chunki = parts * (int)(i / parts)
if (j in range(i-10, i+10)) or (j > chunki and j < (chunki)+(parts/4)) or (j > (chunki)+parts-(parts/4) and j < (chunki)+parts):
simMat[i][j] = 0
continue
a = np.frombuffer(datalist[i], np.int16)
b = np.frombuffer(datalist[j], np.int16)
doublechunk = chunksize * 2
halfchunk = chunksize/2
with warnings.catch_warnings():
warnings.simplefilter("ignore")
apre = copy.deepcopy(a[:(chunksize*jumpsize):jumpsize]).astype(float)
a1 = np.zeros(doublechunk)
a1[halfchunk:halfchunk+chunksize] = apre
bpre = copy.deepcopy(b[:(chunksize*jumpsize):jumpsize]).astype(float)
b1 = np.zeros(doublechunk)
b1[halfchunk:halfchunk+chunksize] = bpre
#automax = np.max(np.correlate(a1, a1, mode='full')[(chunksize/2):])
c = signal.fftconvolve(b1, apre[::-1], mode='valid')
automax = np.max(c)
#compmax = np.max(np.correlate(a1, b1, mode='full')[(chunksize/2):])
d = signal.fftconvolve(a1, apre[::-1], mode='valid')
compmax = np.max(d)
apreJ = copy.deepcopy(a[:(chunksize*jumpsize2):jumpsize2]).astype(float)
a1J = np.zeros(doublechunk)
a1J[halfchunk:halfchunk+chunksize] = apreJ
bpreJ = copy.deepcopy(b[:(chunksize*jumpsize2):jumpsize2]).astype(float)
b1J = np.zeros(doublechunk)
b1J[halfchunk:halfchunk+chunksize] = bpreJ
#automax = np.max(np.correlate(a1, a1, mode='full')[(chunksize/2):])
cJ = signal.fftconvolve(b1J, apreJ[::-1], mode='valid')
automax2 = np.max(cJ)
#compmax = np.max(np.correlate(a1, b1, mode='full')[(chunksize/2):])
dJ = signal.fftconvolve(a1J, apreJ[::-1], mode='valid')
compmax2 = np.max(dJ)
similarity = ((compmax / automax) + (compmax2 / automax2)) / 2
#print (automax)
#print (compmax)
#print (similarity)
if similarity <= 1 and similarity > 0:
simMat[i][j] = similarity
else:
simMat[i][j] = 0
# Requires inport of matplot lib as plt
#plt.plot(npf.ifft(npf.fft(a1)) * npf.fft(a1))
#plt.plot(npf.ifft(npf.fft(a1)) * npf.fft(b1))
auto = np.correlate(a1, a1, mode='same')
other = np.correlate(a1, b1, mode='same')
#plt.plot(auto)
#plt.plot(other)
#plt.show()
#rebuiltdata += bytes(a)
bar.complete()
print ("CALCULATED")
#print (simMat)
indexArr = copy.deepcopy(simMat)
maxIndices = np.argpartition(simMat, -4)[-4:]
simMat = simMat.ravel()
simMat.sort()
simMat = simMat[::-1]
top = simMat[:(20*len(rawdatas))]
# print (top)
test = []
for i in range(len(top)):
test.append(np.asarray(np.where(indexArr == top[i])).T[0].tolist())
fragList = []
fragDict = {test[0][0]: test[0][1]}
for i in range(len(test)):
fragList.append(datalist[test[i][0]] + datalist[test[i][1]])
fragDict[test[i][0]] = test[i][1]
return fragDict, datalist
def get_centroid(img, mask):
correlation = signal.fftconvolve(img,mask,mode='same')
#Getting max
cy, cx = np.unravel_index(correlation.argmax(),correlation.shape)
return cy,cx
def compareAll(self,
rawdatas,
parts=200,
chunksize=100,
threshold=1.0,
maxStop=None,
jumpsize=10,
jumpsize2=100):
if maxStop == None:
maxStop = len(rawdatas) * 500
datalist = []
for rawdata in rawdatas:
datasize = int((int(len(rawdata) / parts)) / 2) * 2
datalist.append(rawdata[0:datasize])
for i in range(1, parts):
datalist.append(rawdata[(datasize * (i)):(datasize * (i + 1))])
rebuiltdata = b''
simMat = np.zeros((len(datalist), len(datalist)))
print("CALCULATING")
bar = progress_bar(len(datalist))
for i in range(len(datalist)):
bar.update_bar(i)
for j in range(len(datalist)):
if i > j:
simMat[i][j] = 0
continue
chunki = parts * (int)(i / parts)
if (j in range(i - 10, i +
10)) or (j > chunki and j < (chunki) +
(parts / 4)) or (j > (chunki) + parts -
(parts / 4) and j <
(chunki) + parts):
simMat[i][j] = 0
continue
a = np.frombuffer(datalist[i], np.int16)
b = np.frombuffer(datalist[j], np.int16)
doublechunk = chunksize * 2
halfchunk = chunksize / 2
with warnings.catch_warnings():
warnings.simplefilter("ignore")
apre = copy.deepcopy(a[:(chunksize *
jumpsize):jumpsize]).astype(float)
a1 = np.zeros(doublechunk)
a1[halfchunk:halfchunk + chunksize] = apre
bpre = copy.deepcopy(b[:(chunksize *
jumpsize):jumpsize]).astype(float)
b1 = np.zeros(doublechunk)
b1[halfchunk:halfchunk + chunksize] = bpre
#automax = np.max(np.correlate(a1, a1, mode='full')[(chunksize/2):])
c = signal.fftconvolve(b1, apre[::-1], mode='valid')
automax = np.max(c)
#compmax = np.max(np.correlate(a1, b1, mode='full')[(chunksize/2):])
d = signal.fftconvolve(a1, apre[::-1], mode='valid')
compmax = np.max(d)
apreJ = copy.deepcopy(
a[:(chunksize * jumpsize2):jumpsize2]).astype(float)
a1J = np.zeros(doublechunk)
a1J[halfchunk:halfchunk + chunksize] = apreJ
bpreJ = copy.deepcopy(
b[:(chunksize * jumpsize2):jumpsize2]).astype(float)
b1J = np.zeros(doublechunk)
b1J[halfchunk:halfchunk + chunksize] = bpreJ
#automax = np.max(np.correlate(a1, a1, mode='full')[(chunksize/2):])
cJ = signal.fftconvolve(b1J, apreJ[::-1], mode='valid')
automax2 = np.max(cJ)
#compmax = np.max(np.correlate(a1, b1, mode='full')[(chunksize/2):])
dJ = signal.fftconvolve(a1J, apreJ[::-1], mode='valid')
compmax2 = np.max(dJ)
similarity = ((compmax / automax) +
(compmax2 / automax2)) / 2
#print (automax)
#print (compmax)
#print (similarity)
if similarity <= 1 and similarity > 0:
simMat[i][j] = similarity
else:
simMat[i][j] = 0
# Requires inport of matplot lib as plt
#plt.plot(npf.ifft(npf.fft(a1)) * npf.fft(a1))
#plt.plot(npf.ifft(npf.fft(a1)) * npf.fft(b1))
auto = np.correlate(a1, a1, mode='same')
other = np.correlate(a1, b1, mode='same')
#plt.plot(auto)
#plt.plot(other)
#plt.show()
#rebuiltdata += bytes(a)
bar.complete()
print("CALCULATED")
#print (simMat)
indexArr = copy.deepcopy(simMat)
maxIndices = np.argpartition(simMat, -4)[-4:]
simMat = simMat.ravel()
simMat.sort()
simMat = simMat[::-1]
top = simMat[:(20 * len(rawdatas))]
# print (top)
test = []
for i in range(len(top)):
test.append(np.asarray(np.where(indexArr == top[i])).T[0].tolist())
fragList = []
fragDict = {test[0][0]: test[0][1]}
for i in range(len(test)):
fragList.append(datalist[test[i][0]] + datalist[test[i][1]])
fragDict[test[i][0]] = test[i][1]
return fragDict, datalist
