def Logistic_regression(modelpath=None):
train_data, train_flag, val_data, val_flag, test_data, test_flag = get_data(
)
if modelpath is not None:
lr = joblib.load(modelpath)
else:
lr = LogisticRegression(penalty='l2',
solver='liblinear',
class_weight="balanced")
lr.fit(train_data, train_flag)
joblib.dump(lr, "params/LR/LR.model")
# val
val_output_prob = lr.predict_proba(val_data)
val_output_prob = np.array(val_output_prob)[:, 1]
_, _, thresold = Measure().get_pr_curve(val_flag, val_output_prob)
test_output_proba = lr.predict_proba(test_data)
test_output_proba = np.array(test_output_proba)[:, 1]
test_output = np.zeros(test_output_proba.shape)
test_output[test_output_proba > thresold] = 1
precision = Measure().Precision(test_flag, test_output)
recall = Measure().Recall(test_flag, test_output)
f1 = Measure().F1_score(test_flag, test_output)
acc = Measure().Accuracy(test_flag, test_output)
print "precision:%.2f\nrecall:%.2f\nf1:%.2f\nacc:%.2f\n" % (
precision, recall, f1, acc)
print "auc:%.2f" % roc_auc_score(test_flag, test_output_proba)
def SVM(modelpath=None): #modelpath="params/SVM/svm_balanced.model"):
train_data, train_flag, val_data, val_flag, test_data, test_flag = get_data(
)
train_data = train_data[:40000]
train_flag = train_flag[:40000]
if modelpath is not None:
svm_ = joblib.load(modelpath)
else:
svm_ = SVC(kernel='rbf', probability=True)
svm_.fit(train_data, train_flag)
joblib.dump(svm_, "params/SVM/svm_balanced.model")
#val
val_output_prob = svm_.predict_proba(val_data)
val_output_prob = np.array(val_output_prob)[:, 1]
_, _, thresold = Measure().get_pr_curve(val_flag, val_output_prob)
#test
test_output_prob = svm_.predict_proba(test_data)
test_output_prob = np.array(test_output_prob)[:, 1]
test_output = np.zeros(test_output_prob.shape)
test_output[test_output_prob > thresold] = 1
precision = Measure().Precision(test_flag, test_output)
recall = Measure().Recall(test_flag, test_output)
f1 = Measure().F1_score(test_flag, test_output)
acc = Measure().Accuracy(test_flag, test_output)
print "model:SVM:\nprecision:%.2f\nrecall:%.2f\nf1:%.2f\nacc:%.2f\n" % (
precision, recall, f1, acc)
print "auc:%.2f" % roc_auc_score(test_flag, test_output_prob)
def __init__(self, pattern, period, simulationTime, units):
Measure.__init__(self, period, simulationTime, units)
self.__periodicValues = PeriodicValues(0, period, simulationTime)
self.__prog = re.compile(pattern)
self.__units = units
def __init__(self, config, eventSched, httpRequester, ownAddrFunc, peerId, persister, pInMeasure, pOutMeasure,
peerPool, connBuilder, connListener, connHandler, choker, torrent, torrentIdent, torrentDataPath, version):
##global stuff
self.config = config
self.version = version
self.peerPool = peerPool
self.connBuilder = connBuilder
self.connListener = connListener
self.connHandler = connHandler
self.choker = choker
##own stuff
self.log = Logger('Bt', '%-6s - ', torrentIdent)
self.torrent = torrent
self.torrentIdent = torrentIdent
self.log.debug("Creating object persister")
self.btPersister = BtObjectPersister(persister, torrentIdent)
self.log.debug("Creating measure classes")
self.inRate = Measure(eventSched, 60, [pInMeasure])
self.outRate = Measure(eventSched, 60, [pOutMeasure])
self.inRate.stop()
self.outRate.stop()
self.log.debug("Creating storage class")
self.storage = Storage(self.config, self.btPersister, torrentIdent, self.torrent, torrentDataPath)
self.log.debug("Creating global status class")
self.pieceStatus = PieceStatus(self.torrent.getTotalAmountOfPieces())
self.log.debug("Creating file priority class")
self.filePrio = FilePriority(self.btPersister, self.version, self.pieceStatus, self.storage.getStatus(),
self.torrent, torrentIdent)
self.log.debug("Creating requester class")
self.requester = Requester(self.config, self.torrentIdent, self.pieceStatus, self.storage, self.torrent)
self.log.debug("Creating tracker requester class")
self.trackerRequester = TrackerRequester(self.config, self.btPersister, eventSched, peerId, self.peerPool, ownAddrFunc, httpRequester,
self.inRate, self.outRate, self.storage, self.torrent, self.torrentIdent, self.version)
self.log.debug("Creating superseeding handler class")
self.superSeedingHandler = SuperSeedingHandler(self.torrentIdent, self.btPersister, self.storage.getStatus(), self.pieceStatus)
##callbacks
self.log.debug("Adding callbacks")
self._addCallbacks()
##status
self.state = 'stopped'
self.started = False
self.paused = True
##lock
self.lock = threading.Lock()
def insertMeasureByPosition(self, width, position=None, counter=None):
if position is None:
if self.numStaffs() == 0:
self._addStaff()
position = NotePosition(self.numStaffs() - 1)
staff = self.getStaff(self.numStaffs() - 1)
position.measureIndex = staff.numMeasures()
self._checkStaffIndex(position.staffIndex)
newMeasure = Measure(width)
newMeasure.counter = counter
staff = self.getStaff(position.staffIndex)
staff.insertMeasure(position, newMeasure)
return newMeasure
def insertMeasureByPosition(self, width, position = None, counter = None):
if position is None:
if self.numStaffs() == 0:
self._addStaff()
position = NotePosition(self.numStaffs() - 1)
staff = self.getStaff(self.numStaffs() - 1)
position.measureIndex = staff.numMeasures()
self._checkStaffIndex(position.staffIndex)
newMeasure = Measure(width)
newMeasure.counter = counter
staff = self.getStaff(position.staffIndex)
staff.insertMeasure(position, newMeasure)
return newMeasure
def model_agg_svm(model_path_dir="params/SVM/Agg_svm"):
train_data, train_flag, val_data, val_flag, test_data, test_flag = get_downsampling_data(
)
model_li = []
for i in range(10):
model_name = "svm-%d.model" % i
path = os.path.join(model_path_dir, model_name)
if os.path.exists(path):
svm_ = joblib.load(os.path.join(model_path_dir, model_name))
else:
print "train submodel-%d)" % (i)
svm_ = SVC(kernel='rbf',
probability=True,
class_weight={
0: 0.45,
1: 0.55
})
svm_.fit(train_data[i], train_flag[i])
joblib.dump(svm_, os.path.join(model_path_dir, model_name))
model_li.append(svm_)
#val
val_output_li = []
for i in range(10):
val_output = model_li[i].predict_proba(val_data)
val_output_li.append(val_output)
val_output_li = np.array(val_output_li)
val_output_li = val_output_li.sum(0) / 10
val_output_proba = val_output_li[:, 1]
_, _, thresold = Measure().get_pr_curve(val_flag, val_output_proba)
#test
test_output_li = []
for i in range(10):
test_output = model_li[i].predict_proba(test_data)
test_output_li.append(test_output)
test_output_li = np.array(test_output_li)
test_output_li = test_output_li.sum(0) / 10
test_output_proba = test_output_li[:, 1]
test_output = np.zeros(test_output_proba.shape)
test_output[test_output_proba > thresold] = 1
precision = Measure().Precision(test_flag, test_output)
recall = Measure().Recall(test_flag, test_output)
f1 = Measure().F1_score(test_flag, test_output)
acc = Measure().Accuracy(test_flag, test_output)
print "model:multi-SVM:\nprecision:%.2f\nrecall:%.2f\nf1:%.2f\nacc:%.2f\n" % (
precision, recall, f1, acc)
print "auc:%.2f" % roc_auc_score(test_flag, test_output_proba)
def insertMeasureByIndex(self, width, index=None, counter=None):
if index is None:
index = self.numMeasures()
if self.numStaffs() == 0:
self._addStaff()
staff = self.getStaff(0)
elif index == self.numMeasures():
staff = self.getStaff(-1)
index = staff.numMeasures()
else:
staff, index = self._staffContainingMeasure(index)
newMeasure = Measure(width)
newMeasure.counter = counter
staff.insertMeasure(NotePosition(measureIndex=index), newMeasure)
return newMeasure
def insertMeasureByIndex(self, width, index = None, counter = None):
if index is None:
index = self.numMeasures()
if self.numStaffs() == 0:
self._addStaff()
staff = self.getStaff(0)
elif index == self.numMeasures():
staff = self.getStaff(-1)
index = staff.numMeasures()
else:
staff, index = self._staffContainingMeasure(index)
newMeasure = Measure(width)
newMeasure.counter = counter
staff.insertMeasure(NotePosition(measureIndex = index),
newMeasure)
return newMeasure
def model_agg_lr(model_path_dir="params/LR/Agg_lr"):
train_data, train_flag, val_data, val_flag, test_data, test_flag = get_downsampling_data(
)
model_li = []
for i in range(10):
model_name = "lr-%d.model" % i
path = os.path.join(model_path_dir, model_name)
if os.path.exists(path):
lr = joblib.load(os.path.join(model_path_dir, model_name))
else:
print "train submodel-%d)" % (i)
lr = LogisticRegression(solver='liblinear',
class_weight="balanced")
lr.fit(train_data[i], train_flag[i])
joblib.dump(lr, os.path.join(model_path_dir, model_name))
model_li.append(lr)
# val
val_output_li = []
for i in range(10):
val_output = model_li[i].predict_proba(val_data)
val_output_li.append(val_output)
val_output_li = np.array(val_output_li)
val_output_li = val_output_li.sum(0) / 10
val_output_proba = val_output_li[:, 1]
_, _, thresold = Measure().get_pr_curve(val_flag, val_output_proba)
#test
test_output_li = []
for i in range(10):
test_output = model_li[i].predict_proba(test_data)
test_output_li.append(test_output)
test_output_li = np.array(test_output_li)
test_output_li = test_output_li.sum(0) / 10
test_output_proba = test_output_li[:, 1]
test_output = np.zeros(test_output_proba.shape)
test_output[test_output_proba > thresold] = 1
precision = Measure().Precision(test_flag, test_output)
recall = Measure().Recall(test_flag, test_output)
f1 = Measure().F1_score(test_flag, test_output)
acc = Measure().Accuracy(test_flag, test_output)
print "precision:%.2f\nrecall:%.2f\nf1:%.2f\nacc:%.2f\n" % (
precision, recall, f1, acc)
print "auc:%.2f" % roc_auc_score(test_flag, test_output_proba)
def get_mean(self):
"""
Get mean of distribution with error from delta theorem
:return: Mean as a Measure
"""
self.calc_cent_moments(1, 2)
val = self.raw_moments[1]
err = (self.cent_moments[2] / self.total_counts)**0.5
return Measure(val, err)
def get_cumulant(self, order):
"""
Calculate cumulant of distribution for given order from central moments with error from delta theorem
:param order: Order of cumulant to calculate
:return: Cumulant value and error as Measure object
"""
self.calc_cent_moments(1, 2 * order)
# Maybe write in bell polynomials later
cm = self.cent_moments
n = self.total_counts
if order == 1:
val = self.raw_moments[1]
err = (cm[2] / n)**0.5
elif order == 2:
val = cm[2]
err = pow((cm[4] - pow(cm[2], 2)) / n, 0.5)
elif order == 3:
val = cm[3]
err = ((cm[6] - cm[3]**2 + 9 * cm[2]**3 - 6 * cm[2] * cm[4]) /
n)**0.5
elif order == 4:
val = cm[4] - 3 * cm[2]**2
err = pow((cm[8] - 12 * cm[6] * cm[2] - 8 * cm[5] * cm[3] -
pow(cm[4], 2) + 48 * cm[4] * pow(cm[2], 2) +
64 * pow(cm[3], 2) * cm[2] - 36 * pow(cm[2], 4)) / n,
0.5)
elif order == 5:
val = cm[5] - 10 * cm[2] * cm[3]
err = pow(
(cm[10] - pow(cm[5], 2) - 10 * cm[4] * cm[6] +
900 * pow(cm[2], 5) - 20 * cm[3] * cm[7] - 20 * cm[8] * cm[2]
+ 125 * cm[2] * pow(cm[4], 2) + 200 * cm[4] * pow(cm[3], 2) -
1000 * pow(cm[3] * cm[2], 2) + 160 * cm[6] * pow(cm[2], 2) -
900 * cm[4] * pow(cm[2], 3) + 240 * cm[2] * cm[3] * cm[5]) /
n, 0.5)
elif order == 6:
val = cm[6] - 15 * cm[4] * cm[2] - 10 * cm[3]**2 + 30 * cm[2]**3
err = pow(
(-30 * cm[4] * cm[8] + 510 * cm[4] * cm[2] * cm[6] +
1020 * cm[4] * cm[3] * cm[5] + 405 * cm[8] * pow(cm[2], 2) -
2880 * cm[6] * pow(cm[2], 3) -
9720 * cm[3] * cm[5] * pow(cm[2], 2) - 30 * cm[2] * cm[10] +
840 * cm[2] * cm[3] * cm[7] + 216 * cm[2] * pow(cm[5], 2) -
40 * cm[3] * cm[9] + 440 * cm[6] * pow(cm[3], 2) - 3600 *
pow(cm[2] * cm[4], 2) - 9600 * cm[2] * cm[4] * pow(cm[3], 2) +
13500 * cm[4] * pow(cm[2], 4) +
39600 * pow(cm[2], 3) * pow(cm[3], 2) + cm[12] -
pow(cm[6], 2) - 12 * cm[5] * cm[7] + 225 * pow(cm[4], 3) -
8100 * pow(cm[2], 6) - 400 * pow(cm[3], 4)) / n, 0.5)
else:
print(f'{order}th order cumulant not implemented, returning nan!')
val = float('nan')
err = float('nan')
return Measure(val, err)
def test():
dst = Dataset(split="test")
valdst=Dataset(split="val")
model=Classifier().cuda()
testloader = data.DataLoader(dst, batch_size=1)
valloader = data.DataLoader(valdst, batch_size=1)
model.load_state_dict(torch.load('params/NN/params2.pkl'))
model.eval()
val_flag = []
val_output_prob = []
for i, _data in enumerate(valloader):
feature, flag = _data
feature = Variable(feature).cuda()
output = model(feature)
output_prob = F.softmax(output, dim=1).detach().cpu().numpy()
val_output_prob.append(output_prob[0][1])
val_flag.append(flag.numpy()[0])
val_flag = np.array(val_flag)
val_output_prob = np.array(val_output_prob)
_,_,thresold=Measure().get_pr_curve(val_flag,val_output_prob)
test_flag=[]
test_output_prob=[]
for i, _data in enumerate(testloader):
feature, flag = _data
feature = Variable(feature).cuda()
output = model(feature)
output_prob=F.softmax(output,dim=1).detach().cpu().numpy()
test_output_prob.append(output_prob[0][1])
test_flag.append(flag.numpy()[0])
test_flag=np.array(test_flag)
test_output_prob=np.array(test_output_prob)
test_output = np.zeros(test_output_prob.shape)
test_output[test_output_prob > thresold] = 1
precision = Measure().Precision(test_flag, test_output)
recall = Measure().Recall(test_flag, test_output)
f1 = Measure().F1_score(test_flag, test_output)
acc = Measure().Accuracy(test_flag, test_output)
print "precision:%f\nrecall:%f\nf1:%f\nacc:%f\n" % (precision, recall, f1, acc)
print "auc:%f"%roc_auc_score(test_flag, test_output_prob)
def model_agg_dt(model_path_dir="params/DecisionTree/Agg_DT"):
train_data, train_flag,val_data, val_flag,test_data, test_flag = get_downsampling_data()
model_li=[]
for i in range(10):
model_name="DT-%d.model"%i
path=os.path.join(model_path_dir,model_name)
if os.path.exists(path):
dt = joblib.load(os.path.join(model_path_dir,model_name))
else:
print "train submodel-%d)"%(i)
dt = tree.DecisionTreeClassifier(max_depth=6, class_weight="balanced")
dt.fit(train_data[i], train_flag[i])
joblib.dump(dt, os.path.join(model_path_dir,model_name))
model_li.append(dt)
#val
val_output_li=[]
for i in range(10):
val_output=model_li[i].predict_proba(val_data)
val_output_li.append(val_output)
val_output_li = np.array(val_output_li)
val_output_li = val_output_li.sum(0) / 10
val_output_proba = val_output_li[:, 1]
_,_,thresold=Measure().get_pr_curve(val_flag,val_output_proba)
#test
test_output_li=[]
for i in range(10):
test_output=model_li[i].predict_proba(test_data)
test_output_li.append(test_output)
test_output_li=np.array(test_output_li)
test_output_li=test_output_li.sum(0)/10
test_output_proba = test_output_li[:, 1]
test_output=np.zeros(test_output_proba.shape)
test_output[test_output_proba>thresold]=1
precision = Measure().Precision(test_flag, test_output)
recall = Measure().Recall(test_flag, test_output)
f1 = Measure().F1_score(test_flag, test_output)
acc = Measure().Accuracy(test_flag, test_output)
print "model:multi-SVM:\nprecision:%.2f\nrecall:%.2f\nf1:%.2f\nacc:%.2f\n" % (precision, recall, f1, acc)
print "auc:%.2f" % roc_auc_score(test_flag, test_output_proba)
def Decision_tree():
train_data, train_flag,val_data,val_flag, test_data, test_flag = get_data()
dt = tree.DecisionTreeClassifier(max_depth=6,class_weight="balanced")
dt = dt.fit(train_data, train_flag)
# val
val_output_prob = dt.predict_proba(val_data)
val_output_prob = np.array(val_output_prob)[:, 1]
_, _, thresold = Measure().get_pr_curve(val_flag, val_output_prob)
# test
test_output_prob = dt.predict_proba(test_data)
test_output_prob = np.array(test_output_prob)[:, 1]
test_output = np.zeros(test_output_prob.shape)
test_output[test_output_prob > thresold] = 1
precision = Measure().Precision(test_flag, test_output)
recall = Measure().Recall(test_flag, test_output)
f1 = Measure().F1_score(test_flag, test_output)
acc = Measure().Accuracy(test_flag, test_output)
print "precision:%.2f\nrecall:%.2f\nf1:%.2f\nacc:%.2f\n" % (precision, recall, f1, acc)
print "auc:%.2f"%roc_auc_score(test_flag, test_output_prob)
def get_sd(self):
"""
Get standard deviation of distribution with error from delta theorem
:return: Standard deviation as a Measure
"""
self.calc_cent_moments(1, 4)
val = self.cent_moments[2]**0.5
err = (self.m[4] - 1) * self.cent_moments[2]
if err >= 0:
err = (err / (4 * self.total_counts))**0.5
else:
err = float('nan')
return Measure(val, err)
def parse_activebandwidthreply(jsonData, clientT_observerT, clientT_observerR,
observerT_remoteR, observerR_remoteR):
bucket = ""
for line in jsonData:
if 'local' in line['Keyword']:
isLocal = True
else:
isLocal = False
direction = line["Direction"]
receiver_identity = line["ReceiverIdentity"]
sender_identity = line["SenderIdentity"]
timestamp = line["Timestamp"]
timestamp_d = datetime.datetime.strptime(
timestamp.replace(".000000", ""), '%Y-%m-%d %H:%M:%S')
bandwidthbps = float(line["Bandwidth [bit/s]"])
bandwidthKbps = bandwidthbps / 1024
bandwidthMbps = bandwidthKbps / 1024
keyword = line['Keyword']
measure = Measure(timestamp_d, bandwidthMbps,
keyword.split("_")[3].replace("noise", "").strip())
if ((direction == "Upstream" and sender_identity == "Client") or \
(direction == "Downstream" and receiver_identity == "Client" )) and bool(isLocal)== True:
clientT_observerT.append(measure)
bucket = "clientT_observerT"
elif ((direction == "Upstream" and sender_identity == "Client") or \
(direction == "Downstream" and receiver_identity == "Client" )) and bool(isLocal)== False:
clientT_observerR.append(measure)
bucket = "clientT_observerR"
elif ((direction == "Upstream" and sender_identity == "Observer") or \
(direction == "Downstream" and receiver_identity == "Observer" )) and bool(isLocal)== True:
observerT_remoteR.append(measure)
bucket = "observerT_remoteR_"
elif ((direction == "Upstream" and sender_identity == "Observer") or \
(direction == "Downstream" and receiver_identity == "Observer" )) and bool(isLocal)== False:
observerR_remoteR.append(measure)
bucket = "observerR_remoteR_"
assert bucket != ""
def get_skewness(self):
"""
Get skewness of distribution with error from delta theorem
:return: Skewness as a Measure
"""
self.calc_cent_moments(1, 6)
m = self.m
if self.cent_moments[2] == 0:
val = float('nan')
err = float('nan')
else:
val = self.cent_moments[3] / self.cent_moments[2]**1.5
err = 9 - 6 * m[4] + m[3]**2 * (
35 + 9 * m[4]) / 4 - 3 * m[3] * m[5] + m[6]
if err >= 0:
err = (err / self.total_counts)**0.5
else:
err = float('nan')
return Measure(val, err)
def get_kurtosis(self):
"""
Get kurtosis of distribution with error from delta theorem
:return: Kurtosis as a Measure
"""
self.calc_cent_moments(1, 8)
m = self.m
if self.cent_moments[2] == 0:
val = float('nan')
err = float('nan')
else:
val = self.cent_moments[4] / self.cent_moments[2]**2 - 3
err = -m[4]**2 + 4 * m[4]**3 + 16 * m[3]**2 * (
1 + m[4]) - 8 * m[3] * m[5] - 4 * m[4] * m[6] + m[8]
if err >= 0:
err = (err / self.total_counts)**0.5
else:
err = float('nan')
return Measure(val, err)
def __init__(self):
'''
retrieve the omero image from the id
:param id: id of an image
'''
self.listFile = [os.path.join(dp, f) for dp, dn, filenames in os.walk(os.getcwd()+'/data') for f in filenames if os.path.splitext(f)[1] == '.tif']
QtCore.QObject.__init__(self)
self.parameter=DEFAULT
self.id = self.listFile[0]
logging.info('Started wood')
self.Image = get_image(self.id)
self.Image = self.Image*1.0/self.Image.max()
self.index = npy.array([self.parameter['FirstLine'],self.parameter['FirstLine']+self.parameter['VisualWidth'],\
self.parameter['FirstColumn'],self.parameter['FirstColumn']+self.parameter['VisualHeight']])
self.updateImg()
self.initToShow()
self.measure = Measure()
'''
if len(self.Image.shape)==3:
self.Image=npy.sum(self.Image,axis=2)/3
self.Image = (self.Image-npy.min(self.Image))*255.0/(npy.max(self.Image)-npy.min(self.Image))
self.Image=self.Image.astype(npy.uint8)
self.Image = self.Image[0:2000,0:10000]
self.set_parameter('DEFAULT')
self.mask = get_mask(self.Image,selemMask =self.get_parameter('selemMask'),low_res=self.get_parameter('low_res'))
self.labels,self.Seg=get_label(self.Image,seg=True,min_size_th=self.get_parameter('min_size_th')\
,max_size_th=self.get_parameter('max_size_th')\
,radius=self.get_parameter('radius')\
,p0=self.get_parameter('p0')\
,iter=self.get_parameter('iter')\
,selemSeg=self.get_parameter('selemSeg'))
self.center,self.tree=get_tree_center(self.mask*self.Seg,self.mask*self.labels)
self.cellsRows = list()
self.selected = npy.zeros((self.center.shape[0],1))
self.compute_Row()
'''
logging.info('Finished')
def get_kurt_var(self):
"""
Get kurtosis*variance of distribution with error from delta theorem
:return: Kurtosis*variance as a Measure
"""
self.calc_cent_moments(1, 8)
m = self.m
if self.cent_moments[2] == 0:
val = float('nan')
err = float('nan')
else:
val = self.cent_moments[4] / self.cent_moments[
2] - 3 * self.cent_moments[2]
err = -9 + 6 * m[4] ** 2 + m[4] ** 3 + 8 * m[3] ** 2 * (5 + m[4]) - 8 * m[3] * m[5] + m[4] * (9 - 2 * m[6]) \
- 6 * m[6] + m[8]
if err >= 0:
err = (err * self.cent_moments[2]**2 / self.total_counts)**0.5
else:
err = float('nan')
return Measure(val, err)
def get_k_stat(self, order):
"""
Calculate k statistic of distribution for given order from central moments
!!!! ERRORS ESTIMATED AS SAME AS CUMULANTS VIA DELTA THEOREM !!!!!
:param order: Order of k statistic to calculate
:return: K statistic value and error as Measure object
"""
self.calc_cent_moments(1, 2 * order)
# Maybe try to find analytical formula later
n = self.total_counts
cm = self.cent_moments
if n < order: # Will get /0 errors
print(f'Too few entries ({n}) for {order} order k-statistic!')
return Measure(float('NaN'), float('NaN'))
if order == 1:
val = self.raw_moments[1]
err = (cm[2] / n)**0.5
elif order == 2:
val = n / (n - 1) * cm[2]
err = pow((cm[4] - pow(cm[2], 2)) / n, 0.5)
elif order == 3:
val = n**2 / ((n - 1) * (n - 2)) * cm[3]
err = ((cm[6] - cm[3]**2 + 9 * cm[2]**3 - 6 * cm[2] * cm[4]) /
n)**0.5
elif order == 4:
val = n**2 * ((n + 1) * cm[4] - 3 *
(n - 1) * cm[2]**2) / ((n - 1) * (n - 2) * (n - 3))
err = pow((cm[8] - 12 * cm[6] * cm[2] - 8 * cm[5] * cm[3] -
pow(cm[4], 2) + 48 * cm[4] * pow(cm[2], 2) +
64 * pow(cm[3], 2) * cm[2] - 36 * pow(cm[2], 4)) / n,
0.5)
elif order == 5:
val = n**3 * ((n + 5) * cm[5] - 10 *
(n - 1) * cm[2] * cm[3]) / ((n - 1) * (n - 2) *
(n - 3) * (n - 4))
err = pow(
(cm[10] - pow(cm[5], 2) - 10 * cm[4] * cm[6] +
900 * pow(cm[2], 5) - 20 * cm[3] * cm[7] - 20 * cm[8] * cm[2]
+ 125 * cm[2] * pow(cm[4], 2) + 200 * cm[4] * pow(cm[3], 2) -
1000 * pow(cm[3] * cm[2], 2) + 160 * cm[6] * pow(cm[2], 2) -
900 * cm[4] * pow(cm[2], 3) + 240 * cm[2] * cm[3] * cm[5]) /
n, 0.5)
elif order == 6:
val = n ** 2 * ((n + 1) * (n ** 2 + 15 * n - 4) * cm[6] - 15 * (n - 1) ** 2 * (n + 4) * cm[2] * cm[4]
- 10 * (n - 1) * (n ** 2 - n + 4) * cm[3] ** 2 + 30 * n * (n - 1) * (n - 2) * cm[2] ** 3) \
/ ((n - 1) * (n - 2) * (n - 3) * (n - 4) * (n - 5))
err = pow(
(-30 * cm[4] * cm[8] + 510 * cm[4] * cm[2] * cm[6] +
1020 * cm[4] * cm[3] * cm[5] + 405 * cm[8] * pow(cm[2], 2) -
2880 * cm[6] * pow(cm[2], 3) -
9720 * cm[3] * cm[5] * pow(cm[2], 2) - 30 * cm[2] * cm[10] +
840 * cm[2] * cm[3] * cm[7] + 216 * cm[2] * pow(cm[5], 2) -
40 * cm[3] * cm[9] + 440 * cm[6] * pow(cm[3], 2) - 3600 *
pow(cm[2] * cm[4], 2) - 9600 * cm[2] * cm[4] * pow(cm[3], 2) +
13500 * cm[4] * pow(cm[2], 4) +
39600 * pow(cm[2], 3) * pow(cm[3], 2) + cm[12] -
pow(cm[6], 2) - 12 * cm[5] * cm[7] + 225 * pow(cm[4], 3) -
8100 * pow(cm[2], 6) - 400 * pow(cm[3], 4)) / n, 0.5)
else:
print(
f'{order}th order k-statistic not implemented, returning nan!')
val = float('nan')
err = float('nan')
return Measure(val, err)
class Bt:
def __init__(self, config, eventSched, httpRequester, ownAddrFunc, peerId, persister, pInMeasure, pOutMeasure,
peerPool, connBuilder, connListener, connHandler, choker, torrent, torrentIdent, torrentDataPath, version):
##global stuff
self.config = config
self.version = version
self.peerPool = peerPool
self.connBuilder = connBuilder
self.connListener = connListener
self.connHandler = connHandler
self.choker = choker
##own stuff
self.log = Logger('Bt', '%-6s - ', torrentIdent)
self.torrent = torrent
self.torrentIdent = torrentIdent
self.log.debug("Creating object persister")
self.btPersister = BtObjectPersister(persister, torrentIdent)
self.log.debug("Creating measure classes")
self.inRate = Measure(eventSched, 60, [pInMeasure])
self.outRate = Measure(eventSched, 60, [pOutMeasure])
self.inRate.stop()
self.outRate.stop()
self.log.debug("Creating storage class")
self.storage = Storage(self.config, self.btPersister, torrentIdent, self.torrent, torrentDataPath)
self.log.debug("Creating global status class")
self.pieceStatus = PieceStatus(self.torrent.getTotalAmountOfPieces())
self.log.debug("Creating file priority class")
self.filePrio = FilePriority(self.btPersister, self.version, self.pieceStatus, self.storage.getStatus(),
self.torrent, torrentIdent)
self.log.debug("Creating requester class")
self.requester = Requester(self.config, self.torrentIdent, self.pieceStatus, self.storage, self.torrent)
self.log.debug("Creating tracker requester class")
self.trackerRequester = TrackerRequester(self.config, self.btPersister, eventSched, peerId, self.peerPool, ownAddrFunc, httpRequester,
self.inRate, self.outRate, self.storage, self.torrent, self.torrentIdent, self.version)
self.log.debug("Creating superseeding handler class")
self.superSeedingHandler = SuperSeedingHandler(self.torrentIdent, self.btPersister, self.storage.getStatus(), self.pieceStatus)
##callbacks
self.log.debug("Adding callbacks")
self._addCallbacks()
##status
self.state = 'stopped'
self.started = False
self.paused = True
##lock
self.lock = threading.Lock()
##internal functions - callbacks
def _addCallbacks(self):
ownStatus = self.storage.getStatus()
self.persistentStatusCallback = self.config.addCallback((('storage', 'persistPieceStatus'),), ownStatus.enablePersisting)
def _removeCallbacks(self):
self.config.removeCallback(self.persistentStatusCallback)
##internal functions - start/pause/stop - common
def _halt(self, targetState):
if self.paused and targetState in ('shutdown', 'remove'):
#stopping and already paused, only need to stop the tracker requester and the callbacks
self.log.debug("Removing callbacks")
self._removeCallbacks()
self.log.debug("Stopping tracker requester")
self.trackerRequester.stop()
else:
#either stopping, removing or shutdown and still running or loading
self.log.debug("Aborting storage loading just in case")
self.storage.abortLoad()
if self.started:
#were already running
self.started = False
if targetState == 'stop':
self.log.debug("Pausing tracker requester")
self.trackerRequester.pause()
else:
self.log.debug("Removing callbacks")
self._removeCallbacks()
self.log.debug("Stopping tracker requester")
self.trackerRequester.stop()
self.log.debug("Removing us from choker")
self.choker.removeTorrent(self.torrentIdent)
self.log.debug("Removing us from connection builder")
self.connBuilder.removeTorrent(self.torrentIdent)
self.log.debug("Removing us from connection listener")
self.connListener.removeTorrent(self.torrent.getTorrentHash())
self.log.debug("Removing us from connection handler")
self.connHandler.removeTorrent(self.torrentIdent)
self.log.debug("Stopping transfer measurement")
self.inRate.stop()
self.outRate.stop()
#shutdown/removal specific tasks which need to be done regardless of current status
if targetState in ('shutdown', 'remove'):
self.log.debug("Removing all infos related to us from connection pool")
self.peerPool.clear(self.torrentIdent)
if targetState == 'remove':
self.log.debug('Removing all persisted objects of this torrent')
self.btPersister.removeAll()
##internal functions - start/pause/stop - specific
def _start(self, loadSuccess):
try:
if loadSuccess:
#loading was successful, add to handlers
self.log.debug("Reseting requester")
self.requester.reset()
self.log.debug("Starting transfer measurement")
self.inRate.start()
self.outRate.start()
self.log.debug("Adding us to connection handler")
self.connHandler.addTorrent(self.torrentIdent, self.torrent, self.pieceStatus, self.inRate, self.outRate, self.storage, self.filePrio, self.requester, self.superSeedingHandler)
self.log.debug("Adding us to connection listener")
self.connListener.addTorrent(self.torrentIdent, self.torrent.getTorrentHash())
self.log.debug("Adding us to connection builder")
self.connBuilder.addTorrent(self.torrentIdent, self.torrent.getTorrentHash())
self.log.debug("Adding us to choker")
self.choker.addTorrent(self.torrentIdent, self.storage.getStatus(), self.superSeedingHandler)
self.log.debug("Starting tracker requester")
self.trackerRequester.start()
self.started = True
self.state = 'running'
except:
#something failed - hard
self.log.error("Error in load function:\n%s", logTraceback())
##external functions - state
def start(self):
#called when torrent is started
self.lock.acquire()
if self.paused:
self.paused = False
if self.storage.isLoaded():
self.log.debug("Storage already loaded, skipping hashing")
self._start(True)
else:
self.storage.load(self._start)
self.state = 'loading'
self.lock.release()
def stop(self):
#called when torrent is stopped
self.lock.acquire()
if not self.paused:
self._halt('stop')
self.paused = True
self.state = 'stopped'
self.lock.release()
def shutdown(self):
#called on shutdown
self.lock.acquire()
self._halt('shutdown')
self.paused = False
self.state = 'stopped'
self.lock.release()
def remove(self):
#called when torrent is removed
self.lock.acquire()
self._halt('remove')
self.paused = False
self.state = 'stopped'
self.lock.release()
##external functions - stats
def getStats(self, wantedStats):
self.lock.acquire()
stats = {}
if wantedStats.get('state', False):
stats['state'] = self.state
#connections
if wantedStats.get('connections', False):
stats.update(self.connHandler.getStats(self.torrentIdent, connDetails=True))
#files
if wantedStats.get('files', False):
stats['files'] = self.filePrio.getStats()
#peers
if wantedStats.get('peers', False) or wantedStats.get('connectionAverages', False):
#get peer stats
connAverages = wantedStats.get('connectionAverages', False)
stats.update(self.peerPool.getStats(self.torrentIdent))
stats.update(self.connHandler.getStats(self.torrentIdent, connSummary=True, connAverages=connAverages))
stats.update(self.trackerRequester.getStats(trackerSummary=True))
#normalise peer stats
if stats['connectedLeeches'] > stats['knownLeeches']:
stats['knownLeeches'] = stats['connectedLeeches']
if stats['connectedSeeds'] > stats['knownSeeds']:
stats['knownSeeds'] = stats['connectedSeeds']
if stats['knownLeeches'] + stats['knownSeeds'] > stats['knownPeers']:
stats['knownPeers'] = stats['knownLeeches'] + stats['knownSeeds']
elif stats['knownLeeches'] + stats['knownSeeds'] < stats['knownPeers']:
stats['knownLeeches'] += stats['knownPeers'] - stats['knownSeeds']
#generate additional conn stats if necessary
if connAverages:
if stats['knownSeeds'] == 0:
stats['knownLeechesPerSeed'] = 0
else:
stats['knownLeechesPerSeed'] = (stats['knownLeeches'] * 1.0) / stats['knownSeeds']
#pieces
if wantedStats.get('pieceAverages', False):
stats.update(self.pieceStatus.getStats(pieceAverages=True))
#progress stats
if wantedStats.get('progress', False):
stats.update(self.storage.getStats())
#requests
if wantedStats.get('requests', False) or wantedStats.get('pieceAverages', False):
reqDetails = wantedStats.get('requests', False)
pieceAverages = wantedStats.get('pieceAverages', False)
stats.update(self.connHandler.getRequesterStats(self.torrentIdent, requestDetails=reqDetails, pieceAverages=pieceAverages))
#tracker
if wantedStats.get('tracker', False):
stats.update(self.trackerRequester.getStats(trackerDetails=True))
if wantedStats.get('trackerStatus', False):
stats.update(self.trackerRequester.getStats(trackerStatus=True))
#transfer stats
if wantedStats.get('transfer', False):
stats['inRawBytes'] = self.inRate.getTotalTransferedBytes()
stats['outRawBytes'] = self.outRate.getTotalTransferedBytes()
stats['inPayloadBytes'] = self.inRate.getTotalTransferedPayloadBytes()
stats['outPayloadBytes'] = self.outRate.getTotalTransferedPayloadBytes()
stats['inRawSpeed'] = self.inRate.getCurrentRate()
stats['outRawSpeed'] = self.outRate.getCurrentRate()
stats['protocolOverhead'] = (100.0 * (stats['inRawBytes'] + stats['outRawBytes'] - stats['inPayloadBytes'] - stats['outPayloadBytes'])) / max(stats['inPayloadBytes'] + stats['outPayloadBytes'], 1.0)
if wantedStats.get('transferAverages', False):
stats['avgInRawSpeed'] = self.inRate.getAverageRate() * 1024
stats['avgOutRawSpeed'] = self.outRate.getAverageRate() * 1024
stats['avgInPayloadSpeed'] = self.inRate.getAveragePayloadRate() * 1024
stats['avgOutPayloadSpeed'] = self.outRate.getAveragePayloadRate() * 1024
#torrent stats
if wantedStats.get('torrent', False):
stats.update(self.torrent.getStats())
stats['superSeeding'] = self.superSeedingHandler.isEnabled()
self.lock.release()
return stats
##external funcs - actions
def setFilePriority(self, fileIds, priority):
self.lock.acquire()
for fileId in fileIds:
self.filePrio.setFilePriority(fileId, priority)
self.lock.release()
def setFileWantedFlag(self, fileIds, wanted):
self.lock.acquire()
if self.started:
#already running, need to go through the connection handler because of syncing issues
self.connHandler.setFileWantedFlag(self.torrentIdent, fileIds, wanted)
else:
#not running
for fileId in fileIds:
self.filePrio.setFileWantedFlag(fileId, wanted)
self.lock.release()
def setSuperSeeding(self, enabled):
self.lock.acquire()
if not enabled == self.superSeedingHandler.isEnabled():
if self.started:
self.connHandler.setSuperSeeding(self.torrentIdent, enabled)
else:
self.superSeedingHandler.setEnabled(enabled)
self.lock.release()
##external funcs - tracker actions
def getTrackerInfo(self):
self.lock.acquire()
trackerInfo = self.trackerRequester.getTrackerInfo()
self.lock.release()
return trackerInfo
def setTrackerInfo(self, newTrackerInfo):
self.lock.acquire()
self.trackerRequester.setTrackerInfo(newTrackerInfo)
self.lock.release()
##external funcs - other
def getInfohash(self):
self.lock.acquire()
infohash = self.torrent.getTorrentHash()
self.lock.release()
return infohash
def __init__(
self,
connStatus,
scheduler,
conn,
direction,
remotePeerAddr,
inMeasure,
outMeasure,
inMeasureParent,
outMeasureParent,
outLimiter,
inLimiter,
msgLenFunc,
msgDecodeFunc,
msgLengthFieldLen,
maxMsgLength,
keepaliveMsgFunc,
log,
):
self.sched = scheduler
# connection
self.conn = conn
self.connIdent = self.conn.fileno()
self.connectTime = time()
self.direction = direction
self.inMsgCount = 0
self.closed = False
# peer
self.remotePeerAddr = remotePeerAddr
# conn status
self.connStatus = connStatus
self.connStatus.addConn(self.connIdent)
# limiter
self.outLimiter = outLimiter
self.outLimiter.addUser(self.connIdent, callback=self.connStatus.allowedToSend, callbackArgs=[self.connIdent])
self.inLimiter = inLimiter
self.inLimiter.addUser(self.connIdent, callback=self.connStatus.allowedToRecv, callbackArgs=[self.connIdent])
# rate
if inMeasure is None:
self.inRate = Measure(self.sched, 60, [inMeasureParent])
else:
self.inRate = inMeasure
self.inRate.start()
if outMeasure is None:
self.outRate = Measure(self.sched, 60, [outMeasureParent])
else:
self.outRate = outMeasure
self.outRate.start()
# data buffer
self.inBuffer = []
self.outBufferQueue = deque()
self.outBufferMessages = {}
self.outBufferMessageId = 0
# messages
self.msgLenFunc = msgLenFunc
self.msgDecodeFunc = msgDecodeFunc
self.msgLengthFieldLen = msgLengthFieldLen
self.maxMsgLength = maxMsgLength
# log
self.log = log
# lock
self.lock = threading.RLock()
# events
self.sendTimeoutEvent = self.sched.scheduleEvent(self.timeout, timedelta=300, funcArgs=["send timed out"])
self.recvTimeoutEvent = self.sched.scheduleEvent(self.timeout, timedelta=300, funcArgs=["read timed out"])
self.keepaliveEvent = self.sched.scheduleEvent(
self.send, timedelta=100, funcArgs=[keepaliveMsgFunc()], repeatdelta=100
)
class Connection:
def __init__(
self,
connStatus,
scheduler,
conn,
direction,
remotePeerAddr,
inMeasure,
outMeasure,
inMeasureParent,
outMeasureParent,
outLimiter,
inLimiter,
msgLenFunc,
msgDecodeFunc,
msgLengthFieldLen,
maxMsgLength,
keepaliveMsgFunc,
log,
):
self.sched = scheduler
# connection
self.conn = conn
self.connIdent = self.conn.fileno()
self.connectTime = time()
self.direction = direction
self.inMsgCount = 0
self.closed = False
# peer
self.remotePeerAddr = remotePeerAddr
# conn status
self.connStatus = connStatus
self.connStatus.addConn(self.connIdent)
# limiter
self.outLimiter = outLimiter
self.outLimiter.addUser(self.connIdent, callback=self.connStatus.allowedToSend, callbackArgs=[self.connIdent])
self.inLimiter = inLimiter
self.inLimiter.addUser(self.connIdent, callback=self.connStatus.allowedToRecv, callbackArgs=[self.connIdent])
# rate
if inMeasure is None:
self.inRate = Measure(self.sched, 60, [inMeasureParent])
else:
self.inRate = inMeasure
self.inRate.start()
if outMeasure is None:
self.outRate = Measure(self.sched, 60, [outMeasureParent])
else:
self.outRate = outMeasure
self.outRate.start()
# data buffer
self.inBuffer = []
self.outBufferQueue = deque()
self.outBufferMessages = {}
self.outBufferMessageId = 0
# messages
self.msgLenFunc = msgLenFunc
self.msgDecodeFunc = msgDecodeFunc
self.msgLengthFieldLen = msgLengthFieldLen
self.maxMsgLength = maxMsgLength
# log
self.log = log
# lock
self.lock = threading.RLock()
# events
self.sendTimeoutEvent = self.sched.scheduleEvent(self.timeout, timedelta=300, funcArgs=["send timed out"])
self.recvTimeoutEvent = self.sched.scheduleEvent(self.timeout, timedelta=300, funcArgs=["read timed out"])
self.keepaliveEvent = self.sched.scheduleEvent(
self.send, timedelta=100, funcArgs=[keepaliveMsgFunc()], repeatdelta=100
)
##internal functions - socket
def _recv(self):
# really recv data - or at least try to
msgs = []
self.sched.rescheduleEvent(self.recvTimeoutEvent, timedelta=300)
wantedBytes = self.conn.getUsedInBufferSpace()
allowedBytes = self.inLimiter.claimUnits(self.connIdent, wantedBytes)
if not allowedBytes == 0:
# may receive something, recv data
data = self.conn.recv(allowedBytes)
self.inRate.updateRate(len(data))
self.inBuffer.append(data)
data = "".join(self.inBuffer)
# process data
msgLen = self.msgLenFunc(data)
while msgLen is not None:
msgLen += self.msgLengthFieldLen # because of the length field
if msgLen > self.maxMsgLength:
# way too large
self._fail("message from peer exceeds size limit (%i bytes)" % (self.maxMsgLength,))
msgLen = None
elif len(data) < msgLen:
# incomplete message
msgLen = None
else:
# finished a message
msg = self.msgDecodeFunc(data[:msgLen])
self._gotMessage(msg)
msgs.append((self.inMsgCount, msg))
self.inMsgCount += 1
data = data[msgLen:]
msgLen = self.msgLenFunc(data)
if data == "":
# all processed
self.inBuffer = []
else:
# still data left
self.inBuffer = [data]
return msgs
def _gotMessage(self, msg):
pass
def _send(self):
# really send the buffered data - or at least try to
self.sched.rescheduleEvent(self.sendTimeoutEvent, timedelta=300)
self.sched.rescheduleEvent(self.keepaliveEvent, timedelta=100)
while len(self.outBufferQueue) > 0:
messageId = self.outBufferQueue[0]
if not messageId in self.outBufferMessages:
# message send got aborted
self.outBufferQueue.popleft()
if len(self.outBufferQueue) == 0:
# nothing to send anymore, notify
self.connStatus.wantsToSend(False, self.connIdent)
else:
# something to send
message = self.outBufferMessages[messageId]
messageLen = len(message[1])
wantedBytes = min(messageLen, self.conn.getFreeOutBufferSpace())
allowedBytes = self.outLimiter.claimUnits(self.connIdent, wantedBytes)
if allowedBytes == 0:
# may not even send a single byte ...
break
else:
# at least something may be send
sendBytes = self.conn.send(message[1][:allowedBytes])
self.outRate.updateRate(sendBytes)
message[2] = True
if sendBytes < messageLen:
# but not all was send
message[1] = message[1][sendBytes:]
break
else:
# all was send
self.outBufferQueue.popleft()
del self.outBufferMessages[messageId]
if len(self.outBufferQueue) == 0:
# nothing to send anymore, notify
self.connStatus.wantsToSend(False, self.connIdent)
if message[0] is not None:
# execute
message[0][0](*message[0][1], **message[0][2])
def _queueSend(self, data, sendFinishedFunc=None, sendFinishedArgs=[], sendFinishedKws={}):
if len(self.outBufferQueue) == 0:
# first queued item, notify about send interest
self.connStatus.wantsToSend(True, self.connIdent)
messageId = self.outBufferMessageId
if sendFinishedFunc is None:
sendFinishedHandle = None
else:
sendFinishedHandle = (sendFinishedFunc, sendFinishedArgs, sendFinishedKws)
self.outBufferMessageId += 1
self.outBufferQueue.append(messageId)
self.outBufferMessages[messageId] = [sendFinishedHandle, data, False]
return messageId
def _abortSend(self, messageId):
aborted = False
message = self.outBufferMessages.get(messageId, None)
if message is not None:
# still queued
if not message[2]:
# send did not start up to now
aborted = True
del self.outBufferMessages[messageId]
return aborted
def _fail(self, reason=""):
# cause the conn to fail
self.log.info("Conn failed: %s", reason)
self.conn.close(force=True)
def _close(self):
# set flag
self.closed = True
# close conn, update conn state
self.conn.close(force=True)
self.connStatus.removeConn(self.connIdent)
# stop rate measurement
self.inRate.stop()
self.outRate.stop()
# remove conn from limiter
self.outLimiter.removeUser(self.connIdent)
self.inLimiter.removeUser(self.connIdent)
# clear buffers
self.inBuffer = []
self.outBufferQueue.clear()
self.outBufferMessages.clear()
self.outBufferMessageId = 0
# remove events
self.sched.removeEvent(self.sendTimeoutEvent)
self.sched.removeEvent(self.recvTimeoutEvent)
self.sched.removeEvent(self.keepaliveEvent)
##internal functions - other
def _getPayloadRatio(self):
inPayload = self.inRate.getTotalTransferedPayloadBytes()
outPayload = self.outRate.getTotalTransferedPayloadBytes()
if inPayload == 0 and outPayload == 0:
ratio = 1.0
elif inPayload == 0 and outPayload != 0:
ratio = 1.0 / outPayload
elif inPayload != 0 and outPayload == 0:
ratio = inPayload / 1.0
else:
ratio = inPayload / (outPayload * 1.0)
return ratio
##external functions - socket
def recv(self):
self.lock.acquire()
msgs = []
if not self.closed:
msgs = self._recv()
self.lock.release()
return msgs
def send(self, data):
self.lock.acquire()
if not self.closed:
self._queueSend(data)
self.lock.release()
def sendEvent(self):
self.lock.acquire()
if not self.closed:
self._send()
self.lock.release()
def fileno(self):
self.lock.acquire()
value = self.connIdent
self.lock.release()
return value
def timeout(self, reason):
self.lock.acquire()
if not self.closed:
self._fail(reason)
self.lock.release()
def close(self):
self.lock.acquire()
if not self.closed:
self._close()
self.lock.release()
##external functions - get info
def getInRate(self):
self.lock.acquire()
rate = self.inRate
self.lock.release()
return rate
def getOutRate(self):
self.lock.acquire()
rate = self.outRate
self.lock.release()
return rate
def getPayloadRatio(self):
self.lock.acquire()
ratio = self._getPayloadRatio()
self.lock.release()
return ratio
def getRemotePeerAddr(self):
self.lock.acquire()
value = self.remotePeerAddr
self.lock.release()
return value
def getShortRemotePeerAddr(self):
self.lock.acquire()
value = self.remotePeerAddr[:10]
self.lock.release()
return value
def new_meas(self):
meas = Measure( angles=self.angles,
dists=np.ones(self.M)*self.max_range,
idxs=np.zeros(self.M, dtype=int)+-9999999)
return meas
def __init__(self, period, simulationTime, units, maintainLastValue=False):
Measure.__init__(self, period, simulationTime, units)
self.periodicAvgValues = PeriodicAvgValues(period, simulationTime, maintainLastValue)
self.__units = units
def main():
conf_path = sys.argv[1]
conf = conf_path.split('/')[-1].replace('.yml', '')
model, opt = load_model(conf_path)
lr_dir = opt['dataroot_LR']
hr_dir = opt['dataroot_GT']
lr_paths = fiFindByWildcard(os.path.join(lr_dir, '*.png'))
hr_paths = fiFindByWildcard(os.path.join(hr_dir, '*.png'))
this_dir = os.path.dirname(os.path.realpath(__file__))
test_dir = os.path.join(this_dir, '..', 'results', conf)
print(f"Out dir: {test_dir}")
measure = Measure(use_gpu=False)
fname = f'measure_full.csv'
fname_tmp = fname + "_"
path_out_measures = os.path.join(test_dir, fname_tmp)
path_out_measures_final = os.path.join(test_dir, fname)
if os.path.isfile(path_out_measures_final):
df = pd.read_csv(path_out_measures_final)
elif os.path.isfile(path_out_measures):
df = pd.read_csv(path_out_measures)
else:
df = None
scale = opt['scale']
pad_factor = 2
for lr_path, hr_path, idx_test in zip(lr_paths, hr_paths, range(len(lr_paths))):
lr = imread(lr_path)
hr = imread(hr_path)
# Pad image to be % 2
h, w, c = lr.shape
lq_orig = lr.copy()
lr = impad(lr, bottom=int(np.ceil(h / pad_factor) * pad_factor - h),
right=int(np.ceil(w / pad_factor) * pad_factor - w))
lr_t = t(lr)
heat = opt['heat']
if df is not None and len(df[(df['heat'] == heat) & (df['name'] == idx_test)]) == 1:
continue
sr_t = model.get_sr(lq=lr_t, heat=heat)
sr = rgb(torch.clamp(sr_t, 0, 1))
sr = sr[:h * scale, :w * scale]
path_out_sr = os.path.join(test_dir, "{:0.2f}".format(heat).replace('.', ''), "{:06d}.png".format(idx_test))
imwrite(path_out_sr, sr)
meas = OrderedDict(conf=conf, heat=heat, name=idx_test)
meas['PSNR'], meas['SSIM'], meas['LPIPS'] = measure.measure(sr, hr)
lr_reconstruct_rgb = imresize(sr, 1 / opt['scale'])
meas['LRC PSNR'] = psnr(lq_orig, lr_reconstruct_rgb)
str_out = format_measurements(meas)
print(str_out)
df = pd.DataFrame([meas]) if df is None else pd.concat([pd.DataFrame([meas]), df])
df.to_csv(path_out_measures + "_", index=False)
os.rename(path_out_measures + "_", path_out_measures)
df.to_csv(path_out_measures, index=False)
os.rename(path_out_measures, path_out_measures_final)
str_out = format_measurements(df.mean())
print(f"Results in: {path_out_measures_final}")
print('Mean: ' + str_out)
def main():
conf_path = sys.argv[1]
conf = conf_path.split('/')[-1].replace('.yml', '')
model, opt = load_model(conf_path)
lr_dir = opt['dataroot_LR']
# hr_dir = opt['dataroot_GT']
lr_paths = fiFindByWildcard(os.path.join(lr_dir, '*.png'))
# hr_paths = fiFindByWildcard(os.path.join(hr_dir, '*.png'))
# this_dir = os.path.dirname(os.path.realpath(__file__))
# test_dir = os.path.join(this_dir, '..', 'results', conf)
this_dir = opt['path']['root']
test_dir = os.path.join(this_dir, 'results', conf)
print(f"Out dir: {test_dir}")
measure = Measure(use_gpu=False)
fname = f'measure_full.csv'
fname_tmp = fname + "_"
path_out_measures = os.path.join(test_dir, fname_tmp)
path_out_measures_final = os.path.join(test_dir, fname)
if os.path.isfile(path_out_measures_final):
df = pd.read_csv(path_out_measures_final)
elif os.path.isfile(path_out_measures):
df = pd.read_csv(path_out_measures)
else:
df = None
scale = opt['scale']
pad_factor = 2
pbar = tqdm(total=len(lr_paths)*opt['n_sample'])
print(f"Total {len(lr_paths)} input images, each generates {opt['n_sample']} output images.")
for lr_path, idx_test in zip(lr_paths, range(len(lr_paths))):
# for lr_path, hr_path, idx_test in zip(lr_paths, hr_paths, range(len(lr_paths))):
lr = imread(lr_path)
# hr = imread(hr_path)
# Pad image to be % 2
h, w, c = lr.shape
lq_orig = lr.copy()
lr = impad(lr, bottom=int(np.ceil(h / pad_factor) * pad_factor - h),
right=int(np.ceil(w / pad_factor) * pad_factor - w))
lr_t = t(lr)
heat = opt['heat']
if df is not None and len(df[(df['heat'] == heat) & (df['name'] == idx_test)]) == 1:
continue
for r in range(opt['n_sample']):
sr_t = model.get_sr(lq=lr_t, heat=heat)
sr = rgb(torch.clamp(sr_t, 0, 1))
sr = sr[:h * scale, :w * scale]
path_out_sr = os.path.join(test_dir, "{:0.2f}".format(heat).replace('.', ''), "{:06d}_sample{:05d}.png".format(idx_test, r))
imwrite(path_out_sr, sr)
pbar.update(1)
def main():
conf_path = sys.argv[1]
conf = conf_path.split('/')[-1].replace('.yml', '')
model, opt = load_model(conf_path)
lr_dir = opt['dataroot_LR']
hr_dir = opt['dataroot_GT']
lr_paths = fiFindByWildcard(os.path.join(lr_dir, '*.png'))
hr_paths = fiFindByWildcard(os.path.join(hr_dir, '*.png'))
# this_dir = os.path.dirname(os.path.realpath(__file__))
# test_dir = os.path.join(this_dir, '..', 'results', conf)
test_dir = opt['path']['results_root'] + '_' + conf
if not os.path.exists(test_dir):
os.makedirs(test_dir)
print(f"Out dir: {test_dir}")
measure = Measure(use_gpu=False)
fname = f'measure_full.csv'
fname_tmp = fname + "_"
path_out_measures = os.path.join(test_dir, fname_tmp)
path_out_measures_final = os.path.join(test_dir, fname)
if os.path.isfile(path_out_measures_final):
df = pd.read_csv(path_out_measures_final)
elif os.path.isfile(path_out_measures):
df = pd.read_csv(path_out_measures)
else:
df = None
scale = opt['scale']
pad_factor = 2
for lr_path, hr_path, idx_test in zip(lr_paths, hr_paths,
range(len(lr_paths))):
print('>>process %s' % (lr_path))
lr = imread(lr_path)
hr = imread(hr_path)
print('\tlr: ', lr.shape)
print('\thr: ', hr.shape)
_, img_name = os.path.split(lr_path)
# Pad image to be % 2
h, w, c = lr.shape
if h >= 338:
print('\tskip ...')
continue
lq_orig = lr.copy()
lr = impad(lr,
bottom=int(np.ceil(h / pad_factor) * pad_factor - h),
right=int(np.ceil(w / pad_factor) * pad_factor - w))
print('\tlr: ', lr.shape)
lr_t = t(lr)
print('\tlr_t: ', lr_t.shape, type(lr_t))
# heat = opt['heat']
for heat in np.linspace(0, 1, num=11):
# Sample a super-resolution for a low-resolution image
print('\theat: ', heat)
if df is not None and len(
df[(df['heat'] == heat) & (df['name'] == idx_test)]) == 1:
continue
sr_t = model.get_sr(lq=lr_t, heat=heat)
print('\tsr_t: ', sr_t.shape, type(sr_t))
sr = rgb(torch.clamp(sr_t, 0, 1))
sr = sr[:h * scale, :w * scale]
print('\tsr: ', sr.shape, type(sr))
save_name = '%s_%s.png' % (img_name.replace('.png', ''), str(heat))
# path_out_sr = os.path.join(test_dir, "{:0.2f}".format(heat).replace('.', ''), save_name)
path_out_sr = os.path.join(test_dir, save_name)
print('\tsave to %s' % (path_out_sr))
imwrite(path_out_sr, sr)
# add compare
h, w, c = np.shape(hr)
lr_resize = cv2.resize(lr, (w, h))
print('\tlr_resize: ', lr_resize.shape)
# TypeError: Expected Ptr<cv::UMat> for argument 'img'
sr = np.array(sr).astype('uint8')
sr = cv2.cvtColor(np.array(sr), cv2.COLOR_RGB2BGR)
print('\tsr: ', sr.shape, type(sr), np.min(sr[:]), np.max(sr[:]))
hr_copy = np.array(hr).astype('uint8')
hr_copy = cv2.cvtColor(np.array(hr_copy), cv2.COLOR_RGB2BGR)
print('\thr: ', sr.shape, type(hr_copy), np.min(hr_copy[:]),
np.max(hr_copy[:]))
start_x = 100
start_y = 100
font_scale = 5
thickness = 6
#lr = lr_resize[:,:,[2,1,0]]
lr_copy = np.array(lr_resize).astype('uint8')
lr_copy = cv2.cvtColor(np.array(lr_copy), cv2.COLOR_RGB2BGR)
color = (0, 0, 255)
cv2.putText(lr_copy, 'lr', (start_x, start_y),
cv2.FONT_HERSHEY_SIMPLEX, font_scale, color, thickness)
cv2.putText(sr, 'sr', (start_x, start_y), cv2.FONT_HERSHEY_SIMPLEX,
font_scale, color, thickness)
cv2.putText(hr_copy, 'hr', (start_x, start_y),
cv2.FONT_HERSHEY_SIMPLEX, font_scale, color, thickness)
save_name = 'compare_%s_%s.png' % (img_name.replace('.png',
''), str(heat))
# path_out_sr_compare = os.path.join(test_dir, "{:0.2f}".format(heat).replace('.', ''), save_name)
path_out_sr_compare = os.path.join(test_dir, save_name)
compare = np.hstack([lr_copy, sr])
compare = np.hstack([compare, hr_copy])
cv2.imwrite(path_out_sr_compare, compare)
print('\tsave to %s' % (path_out_sr_compare))
meas = OrderedDict(conf=conf, heat=heat, name=idx_test)
meas['PSNR'], meas['SSIM'], meas['LPIPS'] = measure.measure(sr, hr)
lr_reconstruct_rgb = imresize(sr, 1 / opt['scale'])
meas['LRC PSNR'] = psnr(lq_orig, lr_reconstruct_rgb)
str_out = format_measurements(meas)
print(str_out)
df = pd.DataFrame([meas]) if df is None else pd.concat(
[pd.DataFrame([meas]), df])
df.to_csv(path_out_measures + "_", index=False)
os.rename(path_out_measures + "_", path_out_measures)
df.to_csv(path_out_measures, index=False)
os.rename(path_out_measures, path_out_measures_final)
str_out = format_measurements(df.mean())
print(f"Results in: {path_out_measures_final}")
print(f'Mean: ' + str_out)
def measure(name, type):
from Measure import Measure
return Measure(name, type)
def get_mean(xs):
return Measure(np.mean(xs), np.std(xs) / np.sqrt(len(xs)))
