def main(argv=None):
if argv is None:
argv = sys.argv
wu2013GroundTruth = "/Users/Ivan/Files/Data/wu2013"
datasetType = 'wu2013'
runName = './Runs/lambda=0.1_hogANDhist_int_f1.p'
run=loadPickle(runName)
#print run
experimenType='default'
run=run.data[experimenType]
#
vidName='jogging-2'
dataset = Dataset(wu2013GroundTruth, datasetType)
viz=VisualizeExperiment(dataset,run)
#viz.show(vidName, experimenType)
#viz.barplot()
viz.precisionAndSuccessPlot(vidName)
def compareDefaultPlots(wildcard="lambda_SE"):
wu2013GroundTruth = "/Users/Ivan/Files/Data/wu2013"
datasetType = 'wu2013'
dataset = Dataset(wu2013GroundTruth, datasetType)
runsNames = glob.glob("./Runs/" + wildcard + "*.p")
formatSave='pdf'
regexp= re.compile("(.*\/)(.+)(.p)")
#s0_e0.2
#s0_e0.5
#s0.1_e0.3
#s0.2_e0.2
#s0.3_e0.4
d=dict()
runs=list()
for runName in runsNames:
m=re.match(regexp,runName)
name=m.group(2)
print name
run = loadPickle(runName)
run.trackerLabel=runName
run.data['TRE'].data=[]
run.data['SRE'].data=[]
d[runName] = run
runs.append(run)
evaluator = Evaluator(dataset, runs)
evaluator.evaluateSingleTracker(runs[0])
def generateAllVizualiations():
wu2013GroundTruth = "/Users/Ivan/Files/Data/wu2013"
folderForGraphs='./Visualizations/'
datasetType = 'wu2013'
dataset = Dataset(wu2013GroundTruth, datasetType)
runsNames = glob.glob('./Runs/upd=3*.p')
runsNames =['lambda_nonorm_s0_e0.5', 'lambda_nonorm_s0.2_e0.2',
'lambda_nonorm_s0.3_e0.4', 'lambda_nonorm_s0.4_e0.4', 'lambda_nonorm_s0.5_e0.3']
#runsNames =['lambda_SE_s0_e0.2', 'lambda_SE_s0_e0.5',
# 'lambda_SE_s0.1_e0.3', 'lambda_SE_s0.2_e0.2', 'lambda_SE_s0.3_e0.4']
for i in range(0,len(runsNames)):
runsNames[i]="./Runs/"+runsNames[i]+".p"
formatSave='pdf'
regexp= re.compile("(.*\/)(.+)(.p)")
for runName in runsNames:
m=re.match(regexp,runName)
name=m.group(2)
print name
run = loadPickle(runName)
viz = VisualizeAllExperiments(dataset, run)
viz.barplot(savefile=folderForGraphs+name+"."+ formatSave)
def main(argv=None):
if argv is None:
argv = sys.argv
wu2013GroundTruth = "/Users/Ivan/Files/Data/Tracking_benchmark"
datasetType = 'wu2013'
runName = './Runs/fk_hist_int_f1.p'
run=loadPickle(runName)
#print run
experimenType='SRE'
#
vidName='jogging-2'
dataset = Dataset(wu2013GroundTruth, datasetType)
viz=VisualizeExperiment(dataset,run)
viz.show(vidName, experimenType)
viz.barplot()
def createSavedEvaluations(wildcard):
wu2013results = "/Users/Ivan/Files/Results/Tracking/wu2013"
wu2013GroundTruth = "/Users/Ivan/Files/Data/Tracking_benchmark"
vot2014Results = "/Users/Ivan/Files/Results/Tracking/vot2014"
vot2014GrounTruth = "/Users/Ivan/Files/Data/vot2014"
datasetType = 'wu2013'
dataset = Dataset(wu2013GroundTruth, datasetType)
runsNames = glob.glob('./Runs/' + wildcard + '*.p')
runs = list()
#
names = list()
for runName in runsNames:
run = loadPickle(runName)
names.append(runName)
print runName
runs.append(run)
evaluator = EvaluatorAllExperiments(dataset, runs, names)
strSave = './Results/'
evaluator.calculateMetricsAndSave(strSave)
def createSavedEvaluations(wildcard):
wu2013results = "/Users/Ivan/Files/Results/Tracking/wu2013"
wu2013GroundTruth = "/Users/Ivan/Files/Data/wu2013"
vot2014Results = "/Users/Ivan/Files/Results/Tracking/vot2014"
vot2014GrounTruth = "/Users/Ivan/Files/Data/vot2014"
datasetType = 'wu2013'
dataset = Dataset(wu2013GroundTruth, datasetType)
#runsNames =['lambda_gray_s0_e0.4', 'lambda_gray_s0.2_e0.4',
# 'lambda_gray_s0.3_e0.3', 'lambda_gray_s0.4_e0.4', 'lambda_gray_s0.5_e0.3']
runsNames = list()
for i in range(0, len(runsNames)):
runsNames[i] = './Runs/' + runsNames[i] + ".p"
runs = list()
runsNames = glob.glob('./Runs/' + wildcard + '*.p')
#
names = list()
for runName in runsNames:
run = loadPickle(runName)
names.append(runName)
print runName
runs.append(run)
evaluator = EvaluatorAllExperiments(dataset, runs, names)
strSave = './Results/'
evaluator.calculateMetricsAndSave(strSave)
def plot_OPE_comparison():
wu2013GroundTruth = "/Users/Ivan/Files/Data/wu2013"
datasetType = 'wu2013'
dataset = Dataset(wu2013GroundTruth, datasetType)
(runsNames, corrected_names) = getRunsForCVPR2016()
runs = list()
names = list()
for runName in runsNames:
run = loadPickle(runName)
names.append(runName)
runs.append(run)
plotMetricsDict = dict()
completeMetricDict = dict()
alternativeNames = corrected_names
if len(alternativeNames) == 0:
for r in runs:
plotMetricsDict[r.name] = r.plotMetricsDictEntry
completeMetricDict[r.name] = r.completeMetricDictEntry
else:
for r, name in zip(runs, alternativeNames):
plotMetricsDict[name] = r.plotMetricsDictEntry
completeMetricDict[name] = r.completeMetricDictEntry
success = list()
precision = list()
names = list()
import re
for name, value in plotMetricsDict.iteritems():
s = value['default'][5]
p = value['default'][4]
success.append(s)
precision.append(p)
if "ObjStruck with " in name:
short_name = re.sub("ObjStruck with ", '', name)
else:
short_name = name
names.append(short_name)
df = pd.DataFrame({
'name': names,
'success': success,
'precision': precision
})
print df
sns.set_style("whitegrid")
plt.subplots_adjust(bottom=0.4)
ax = sns.barplot(x="name", y="success", data=df, palette="Blues_d")
locs, labels = plt.xticks()
plt.setp(labels, rotation=-45)
plt.show()
def plot_OPE_comparison():
wu2013GroundTruth = "/Users/Ivan/Files/Data/wu2013"
datasetType = 'wu2013'
dataset = Dataset(wu2013GroundTruth, datasetType)
(runsNames, corrected_names)= getRunsForCVPR2016()
runs = list()
names=list()
for runName in runsNames:
run = loadPickle(runName)
names.append(runName)
runs.append(run)
plotMetricsDict=dict()
completeMetricDict=dict()
alternativeNames=corrected_names
if len(alternativeNames) == 0:
for r in runs:
plotMetricsDict[r.name]=r.plotMetricsDictEntry
completeMetricDict[r.name]=r.completeMetricDictEntry
else:
for r,name in zip(runs,alternativeNames):
plotMetricsDict[name]=r.plotMetricsDictEntry
completeMetricDict[name]=r.completeMetricDictEntry
success=list()
precision=list()
names=list()
import re
for name, value in plotMetricsDict.iteritems():
s=value['default'][5]
p=value['default'][4]
success.append(s)
precision.append(p)
if "ObjStruck with " in name:
short_name=re.sub("ObjStruck with ",'',name)
else:
short_name=name
names.append(short_name)
df = pd.DataFrame({'name': names, 'success': success, 'precision':precision})
print df
sns.set_style("whitegrid")
plt.subplots_adjust(bottom=0.4)
ax=sns.barplot(x="name", y="success", data=df,palette="Blues_d")
locs, labels = plt.xticks()
plt.setp(labels, rotation=-45)
plt.show()
def plotStraddelingEdgeOPE(wildcard, savefilename='', format='pdf'):
wu2013GroundTruth = "/Users/Ivan/Files/Data/wu2013"
folderForGraphs='./Visualizations/'
datasetType = 'wu2013'
dataset = Dataset(wu2013GroundTruth, datasetType)
runsNames = glob.glob("./Results/" + wildcard + "*.p")
#runsNames = glob.glob('./Runs/upd=3_hogANDhist_int*.p')
formatSave='pdf'
regexp= re.compile("(.*\/)(.+)(.p)")
experimentType='default'
d=dict()
for runName in runsNames:
m=re.match(regexp,runName)
name=m.group(2)
run = loadPickle(runName)
results = run.plotMetricsDictEntry['default']
d[name]=results[4:6]
(df1, df2)=reshuffleStraddlingEdgeOPEData(d)
save = False
# Initialize a grid of plots with an Axes for each walk
grid = sns.FacetGrid(df1, col="$\lambda_e$", hue="$\lambda_e$", col_wrap=3, size=3)
# Draw a horizontal line to show the starting point
grid.map(plt.axhline, y=0.749, ls=":", c=".5")
# Draw a line plot to show the trajectory of each random walk
grid.map(plt.plot, "$\lambda_s$", "Precision", marker="o", ms=4)
grid.set(xticks=np.arange(6)/float(10), yticks=[0.6, 0.65, 0.7, 0.75, 0.8, 0.85],
xlim=(0, 0.5), ylim=(0.6, 0.85))
grid.fig.tight_layout(w_pad=1)
# Adjust the tick positions and labels
if savefilename=='':
plt.show()
else:
plt.savefig(savefilename+"precision."+format)
grid = sns.FacetGrid(df2, col="$\lambda_e$", hue="$\lambda_e$", col_wrap=3, size=3)
grid.map(plt.axhline, y=0.59, ls=":", c=".5")
grid.map(plt.plot, "$\lambda_s$", "Success", marker="o", ms=4)
grid.set(xticks=np.arange(6)/float(10), yticks=[0.5, 0.55, 0.6, 0.65, 0.7],
xlim=(0, 0.5), ylim=(0.5, 0.7))
grid.fig.tight_layout(w_pad=1)
if savefilename=='':
plt.show()
else:
plt.savefig(savefilename+"success."+format)
def plotOPE_SRE_TRE_Robust(
saveFigureToFolder='/Users/Ivan/Code/Tracking/Antrack/doc/technical_reports/images/',
format='png'):
wu2013results = "/Users/Ivan/Files/Results/Tracking/wu2013"
wu2013GroundTruth = "/Users/Ivan/Files/Data/wu2013"
vot2014Results = "/Users/Ivan/Files/Results/Tracking/vot2014"
vot2014GrounTruth = "/Users/Ivan/Files/Data/vot2014"
datasetType = 'wu2013'
dataset = Dataset(wu2013GroundTruth, datasetType)
(runsNames, corrected_names) = getRunsForCVPR2016_robust()
runs = list()
names = list()
for runName in runsNames:
run = loadPickle(runName)
names.append(runName)
#run = run.data[experimentType]
runs.append(run)
# run=loadPickle('./Runs/TLD.p')
# runs.append(run)
# names.append('./Runs/TLD.p')
save = True
evaluator = EvaluatorAllExperiments(dataset, list(), names)
successAndPrecision = 'cvpr2016_SuccessAndPrecision_wu2013'
histograms = 'cvpr2016_histogram_wu2013'
i = format
if save:
evaluator.evaluateFromSave(
runs,
successAndPrecisionPlotName=saveFigureToFolder +
successAndPrecision + '.' + i,
histogramPlot=saveFigureToFolder + histograms + '.' + i,
alternativeNames=corrected_names)
else:
evaluator.evaluateFromSave(runs, alternativeNames=corrected_names)
def getRuns(self,wildcard):
runsNames = glob.glob(self.folder + wildcard + '*.p')
runs = list()
#
names = list()
for runName in runsNames:
run = loadPickle(runName)
names.append(runName)
# run = run.data[experimentType]
runs.append(run)
plotMetricsDict = dict()
completeMetricDict = dict()
for r in runs:
plotMetricsDict[r.name] = r.plotMetricsDictEntry
completeMetricDict[r.name] = r.completeMetricDictEntry
return (plotMetricsDict,completeMetricDict)
def plotOPE_SRE_TRE_Robust(saveFigureToFolder = '/Users/Ivan/Code/Tracking/Antrack/doc/technical_reports/images/',
format='png'):
wu2013results = "/Users/Ivan/Files/Results/Tracking/wu2013"
wu2013GroundTruth = "/Users/Ivan/Files/Data/wu2013"
vot2014Results = "/Users/Ivan/Files/Results/Tracking/vot2014"
vot2014GrounTruth = "/Users/Ivan/Files/Data/vot2014"
datasetType = 'wu2013'
dataset = Dataset(wu2013GroundTruth, datasetType)
(runsNames, corrected_names)= getRunsForCVPR2016_robust()
runs = list()
names=list()
for runName in runsNames:
run = loadPickle(runName)
names.append(runName)
#run = run.data[experimentType]
runs.append(run)
# run=loadPickle('./Runs/TLD.p')
# runs.append(run)
# names.append('./Runs/TLD.p')
save = True
evaluator = EvaluatorAllExperiments(dataset, list(), names)
successAndPrecision = 'cvpr2016_SuccessAndPrecision_wu2013'
histograms = 'cvpr2016_histogram_wu2013'
i = format
if save:
evaluator.evaluateFromSave(runs,successAndPrecisionPlotName=saveFigureToFolder+successAndPrecision+'.'+
i,histogramPlot=saveFigureToFolder+histograms+'.'+
i, alternativeNames=corrected_names)
else:
evaluator.evaluateFromSave(runs, alternativeNames=corrected_names)
def generateAllVizualiations():
wu2013GroundTruth = "/Users/Ivan/Files/Data/Tracking_benchmark"
folderForGraphs='./Visualizations/'
datasetType = 'wu2013'
dataset = Dataset(wu2013GroundTruth, datasetType)
runsNames = glob.glob('./Runs/upd*.p')
formatSave='pdf'
regexp= re.compile("(.*\/)(.+)(.p)")
for runName in runsNames:
m=re.match(regexp,runName)
name=m.group(2)
print name
run = loadPickle(runName)
viz = VisualizeAllExperiments(dataset, run)
viz.barplot(savefile=folderForGraphs+name+"."+ formatSave)
def plotComparisonToOtherTrackers(dataset, saveFigureToFolders,save):
# TODO: Add TLD tracker and, perhaps, SCM
runsNames = ['SAMF','DSST', 'upd=3_hogANDhist_int_f1','a30_hogANDhist_int_f1']
runs = list()
#
names = list()
for runName in runsNames:
runName = './Results/' + runName + '.p'
run = loadPickle(runName)
names.append(runName)
print runName
# run = run.data[experimentType]
runs.append(run)
# run=loadPickle('./Runs/TLD.p')
# runs.append(run)
# names.append('./Runs/TLD.p')
evaluator = EvaluatorAllExperiments(dataset, list(), names)
# saveFormat = ['png', 'pdf']
saveFormat = ['eps']
successAndPrecision = 'SuccessAndPrecision_wu2013'
histograms = 'histogram_wu2013'
if save:
for folder in saveFigureToFolders:
evaluator.evaluateFromSave(runs,successAndPrecisionPlotName=folder+"/"+successAndPrecision+'.'+
saveFormat[0],histogramPlot=folder+ "/"+histograms+'.'+
saveFormat[0])
else:
evaluator.evaluateFromSave(runs)
def generateDefaultVizualiations(wildcard):
wu2013GroundTruth = "/Users/Ivan/Files/Data/wu2013"
folderForGraphs='./Visualizations/'
datasetType = 'wu2013'
dataset = Dataset(wu2013GroundTruth, datasetType)
runsNames = glob.glob("./Runs/" + wildcard + "*.p")
#runsNames = glob.glob('./Runs/upd=3_hogANDhist_int*.p')
formatSave='pdf'
regexp= re.compile("(.*\/)(.+)(.p)")
for runName in runsNames:
m=re.match(regexp,runName)
name=m.group(2)
print name
run = loadPickle(runName)
run.data['TRE'].data=[]
run.data['SRE'].data=[]
viz = VisualizeAllExperiments(dataset, run)
viz.barplotDefault(savefile=folderForGraphs+name+"."+ formatSave)
def main(argv=None):
if argv is None:
argv = sys.argv
wu2013GroundTruth = "/Users/Ivan/Files/Data/Tracking_benchmark"
datasetType = 'wu2013'
experimentType='default'
runName = './Runs/a28_hist_int_f1.p'
run = loadPickle(runName)
# print run
vidName = 'basketball'
dataset = Dataset(wu2013GroundTruth, datasetType)
viz = VisualizeAllExperiments(dataset, run)
#viz.precisionAndSuccessPlot()
# exp1=run.data['default']
#
# vizOne=VisualizeExperiment(dataset,exp1)
#
# vizOne.precisionAndSuccessPlot(vidName)
#viz.show(vidName,experimentRunNumber=0,experiment='default')
#viz.show(vidName,100)
runsNames = glob.glob('./Runs/*.p')
for r in runsNames:
print r
def plotSensitivitySpecific(paperPlot, baselineRun,wildcards,savefile=''):
plist=list()
slist=list()
xvalsList=list()
for wildcard in wildcards:
(ps,ss, xValuess) = getPrecisionSuccessGivenWildCard(paperPlot, wildcard)
plist.append(ps)
slist.append(ss)
xvalsList.append(xValuess)
run = loadPickle(baselineRun)
baseline_p=run.completeMetricDictEntry[0]
baseline_s = run.completeMetricDictEntry[1]
cm = plt.get_cmap('gist_rainbow')
NUM_COLORS = len(wildcards)+ 1
titleFontSize = paperPlot.titleFontSize;
headerFontSize = paperPlot.headerFontSize;
axisFontSize = paperPlot.axisFontSize;
lineWidth = paperPlot.lineWidth;
legendSize = paperPlot.legendSize;
labelsFontSize = paperPlot.labelsFontSize
minY=paperPlot.minY
maxY=paperPlot.maxY
deltaPrecision=paperPlot.deltaPrecision
# this should be just regular plot: value vs b
# get values from names
max_yticks = 5
# now everything is sorted
fig, ax = plt.subplots(nrows=2, ncols=4, sharex=True, figsize=(13,5))
#plt.suptitle('Sensitivity analysis for parameter: '+wildcard,fontsize=paperPlot.axisFontSize+4)
for xValues, p, i in zip(xvalsList, plist, range(0,len(wildcards))):
#ax[0,i].subplot(2,4,i+1 )
ax[0,i].plot(xValues,p, paperPlot.sensitivityColorAndPoints, linewidth=lineWidth, color=cm(1.*i/NUM_COLORS))
ax[0,i].plot([min(min(xvalsList[0]),0), max(xvalsList[0])], [baseline_p, baseline_p], color='k', linestyle='--', linewidth=lineWidth)
ax[0,i].set_ylim(minY+0.05+ deltaPrecision,maxY+ deltaPrecision-0.05)
title = wildcards[i].replace("=",'')
yloc = plt.MaxNLocator(max_yticks)
if i != 0:
ax[0, i].get_xaxis().set_visible(False)
ax[0,i].get_yaxis().set_visible(False)
ax[0,i].set_title(title.upper())
else:
ax[0,i].set_ylabel('Precision', color='black')
ax[0,i].set_title(title)
ax[0,i].yaxis.set_major_locator(yloc)
#plt.ylim((minY+ deltaPrecision,maxY+ deltaPrecision))
#plt.ylim((minY+ deltaPrecision,maxY+ deltaPrecision))
#ax2 = ax[0,3].twinx()
#ax2.set_ylabel('Precision', color='black')
#plt.xlabel(wildcard, fontsize=axisFontSize)
#plt.ylabel("Precision", fontsize=axisFontSize)
#plt.legend(['filter on','filter off'])
#plt.subplot(2, 4, 5)
for xValues, s, i in zip(xvalsList, slist, range(0,len(wildcards))):
#ax[0,i].subplot(2,4,i+1 )
ax[1,i].plot(xValues,s, paperPlot.sensitivityColorAndPoints, linewidth=lineWidth, color=cm(1.*i/NUM_COLORS))
ax[1,i].plot([min(min(xvalsList[0]),0), max(xvalsList[0])], [baseline_s, baseline_s], color='k', linestyle='--', linewidth=lineWidth)
ax[1,i].set_ylim(minY+0.05,maxY-0.05)
title = wildcards[i].replace("=",'')
yloc = plt.MaxNLocator(max_yticks)
if i != 0:
ax[1,i].get_yaxis().set_visible(False)
else:
ax[1,i].set_ylabel('Success', color='black')
ax[1,i].yaxis.set_major_locator(yloc)
#for xValues, s,i in zip(xvalsList, slist, range(0,len(wildcards))):
# plt.plot(xValues,s, paperPlot.sensitivityColorAndPoints, linewidth=lineWidth, color=cm(1.*i/NUM_COLORS))
#plt.plot([min(min(xvalsList[0]), 0), max(xvalsList[0])], [baseline_s, baseline_s], color='k', linestyle='--',
# linewidth=lineWidth)
#plt.ylim((minY, maxY))
#plt.xlabel(wildcard, fontsize=axisFontSize)
#plt.ylabel("Success", fontsize=axisFontSize)
if savefile == '':
plt.show()
else:
plt.savefig(savefile)
# trackerLabel="STR+f_hog"
# wildcard = sys.argv[1]
dataset = Dataset(wu2013GroundTruth, datasetType)
runsNames = glob.glob('./Results/' + wildcard + '*.p')
#runsNames = ['SAMF', 'Kernelized_filter', 'fk_hist_int_f0', 'fk_hist_int_f1','TLD']
runs = list()
#
names=list()
for runName in runsNames:
#runName= './Results/' + runName + '.p'
run = loadPickle(runName)
names.append(runName)
print runName
#run = run.data[experimentType]
runs.append(run)
# run=loadPickle('./Runs/TLD.p')
# runs.append(run)
# names.append('./Runs/TLD.p')
evaluator = EvaluatorAllExperiments(dataset, list(), names)
saveFigureToFolder = '/Users/Ivan/Code/personal-website/Projects/Object_aware_tracking/images/multiScale/'
#saveFormat = ['png', 'pdf']
saveFormat=['pdf']
# wildcard = sys.argv[1]
dataset = Dataset(wu2013GroundTruth, datasetType)
runsNames = glob.glob('./Results/' + wildcard + '*.p')
runsNames = getRunsForCVPR2016()
#runsNames = ['SAMF', 'Kernelized_filter', 'fk_hist_int_f0', 'fk_hist_int_f1','TLD']
runs = list()
#
names = list()
for runName in runsNames:
#runName= './Results/' + runName + '.p'
run = loadPickle(runName)
names.append(runName)
print runName
#run = run.data[experimentType]
runs.append(run)
# run=loadPickle('./Runs/TLD.p')
# runs.append(run)
# names.append('./Runs/TLD.p')
evaluator = EvaluatorAllExperiments(dataset, list(), names)
#saveFigureToFolder = '/Users/Ivan/Code/personal-website/Projects/Object_aware_tracking/images/multiScale/'
saveFigureToFolder = '/Users/Ivan/Code/Tracking/Antrack/doc/technical_reports/images/'
#saveFormat = ['png', 'pdf']
saveFormat = ['png']
def plotSensitivityUpd(paperPlot, baselineRun, wildcard, savefile=''):
(p, s, xValues) = getPrecisionSuccessGivenWildCard(paperPlot, wildcard)
run = loadPickle(baselineRun)
baseline_p = run.completeMetricDictEntry[0]
baseline_s = run.completeMetricDictEntry[1]
titleFontSize = paperPlot.titleFontSize
headerFontSize = paperPlot.headerFontSize
axisFontSize = paperPlot.axisFontSize
lineWidth = paperPlot.lineWidth
legendSize = paperPlot.legendSize
labelsFontSize = paperPlot.labelsFontSize
minY = paperPlot.minY
maxY = paperPlot.maxY
deltaPrecision = paperPlot.deltaPrecision
cm = plt.get_cmap('gist_rainbow')
NUM_COLORS = 3
max_yticks = 5
# now everything is sorted
fig, ax = plt.subplots(nrows=1, ncols=2, sharex=True, figsize=(5, 4))
fig.subplots_adjust(hspace=5)
#plt.suptitle('Sensitivity analysis for parameter: '+wildcard,fontsize=paperPlot.axisFontSize+4)
i = 0
ax[0].plot(xValues,
p,
paperPlot.sensitivityColorAndPoints,
linewidth=lineWidth,
color=cm(1. * i / NUM_COLORS))
ax[0].plot([min(min(xValues), 1), max(xValues)], [baseline_p, baseline_p],
color='k',
linestyle='--',
linewidth=lineWidth)
ax[0].set_ylim(minY + 0.05 + deltaPrecision, maxY + deltaPrecision - 0.05)
title = wildcard.replace("=", '')
yloc = plt.MaxNLocator(max_yticks)
ax[i].set_ylabel('Precision', color='black')
ax[i].yaxis.set_major_locator(yloc)
#plt.ylim((minY+ deltaPrecision,maxY+ deltaPrecision))
#plt.ylim((minY+ deltaPrecision,maxY+ deltaPrecision))
#ax2 = ax[0,3].twinx()
#ax2.set_ylabel('Precision', color='black')
#plt.xlabel(wildcard, fontsize=axisFontSize)
#plt.ylabel("Precision", fontsize=axisFontSize)
#plt.legend(['filter on','filter off'])
#plt.subplot(2, 4, 5)
ax[1].plot(xValues,
s,
paperPlot.sensitivityColorAndPoints,
linewidth=lineWidth,
color=cm(1. * i / NUM_COLORS))
ax[1].plot([min(min(xValues), 1), max(xValues)], [baseline_s, baseline_s],
color='k',
linestyle='--',
linewidth=lineWidth)
ax[1].set_ylim(minY + 0.05, maxY - 0.05)
title = wildcard.replace("=", '')
yloc = plt.MaxNLocator(max_yticks)
i = 0
ax[1].set_ylabel('Success', color='black')
ax[1].yaxis.set_major_locator(yloc)
#for xValues, s,i in zip(xvalsList, slist, range(0,len(wildcards))):
# plt.plot(xValues,s, paperPlot.sensitivityColorAndPoints, linewidth=lineWidth, color=cm(1.*i/NUM_COLORS))
#plt.plot([min(min(xvalsList[0]), 0), max(xvalsList[0])], [baseline_s, baseline_s], color='k', linestyle='--',
# linewidth=lineWidth)
#plt.ylim((minY, maxY))
#plt.xlabel(wildcard, fontsize=axisFontSize)
#plt.ylabel("Success", fontsize=axisFontSize)
if savefile == '':
plt.show()
else:
plt.savefig(savefile)
def plotSensitivitySpecific(paperPlot, baselineRun, wildcards, savefile=''):
plist = list()
slist = list()
xvalsList = list()
for wildcard in wildcards:
(ps, ss,
xValuess) = getPrecisionSuccessGivenWildCard(paperPlot, wildcard)
plist.append(ps)
slist.append(ss)
xvalsList.append(xValuess)
run = loadPickle(baselineRun)
baseline_p = run.completeMetricDictEntry[0]
baseline_s = run.completeMetricDictEntry[1]
cm = plt.get_cmap('gist_rainbow')
NUM_COLORS = len(wildcards) + 1
titleFontSize = paperPlot.titleFontSize
headerFontSize = paperPlot.headerFontSize
axisFontSize = paperPlot.axisFontSize
lineWidth = paperPlot.lineWidth
legendSize = paperPlot.legendSize
labelsFontSize = paperPlot.labelsFontSize
minY = paperPlot.minY
maxY = paperPlot.maxY
deltaPrecision = paperPlot.deltaPrecision
# this should be just regular plot: value vs b
# get values from names
max_yticks = 5
# now everything is sorted
fig, ax = plt.subplots(nrows=2, ncols=4, sharex=True, figsize=(13, 5))
#plt.suptitle('Sensitivity analysis for parameter: '+wildcard,fontsize=paperPlot.axisFontSize+4)
for xValues, p, i in zip(xvalsList, plist, range(0, len(wildcards))):
#ax[0,i].subplot(2,4,i+1 )
ax[0, i].plot(xValues,
p,
paperPlot.sensitivityColorAndPoints,
linewidth=lineWidth,
color=cm(1. * i / NUM_COLORS))
ax[0, i].plot([min(min(xvalsList[0]), 0),
max(xvalsList[0])], [baseline_p, baseline_p],
color='k',
linestyle='--',
linewidth=lineWidth)
ax[0, i].set_ylim(minY + 0.05 + deltaPrecision,
maxY + deltaPrecision - 0.05)
title = wildcards[i].replace("=", '')
yloc = plt.MaxNLocator(max_yticks)
if i != 0:
ax[0, i].get_xaxis().set_visible(False)
ax[0, i].get_yaxis().set_visible(False)
ax[0, i].set_title(title.upper())
else:
ax[0, i].set_ylabel('Precision', color='black')
ax[0, i].set_title(title)
ax[0, i].yaxis.set_major_locator(yloc)
#plt.ylim((minY+ deltaPrecision,maxY+ deltaPrecision))
#plt.ylim((minY+ deltaPrecision,maxY+ deltaPrecision))
#ax2 = ax[0,3].twinx()
#ax2.set_ylabel('Precision', color='black')
#plt.xlabel(wildcard, fontsize=axisFontSize)
#plt.ylabel("Precision", fontsize=axisFontSize)
#plt.legend(['filter on','filter off'])
#plt.subplot(2, 4, 5)
for xValues, s, i in zip(xvalsList, slist, range(0, len(wildcards))):
#ax[0,i].subplot(2,4,i+1 )
ax[1, i].plot(xValues,
s,
paperPlot.sensitivityColorAndPoints,
linewidth=lineWidth,
color=cm(1. * i / NUM_COLORS))
ax[1, i].plot([min(min(xvalsList[0]), 0),
max(xvalsList[0])], [baseline_s, baseline_s],
color='k',
linestyle='--',
linewidth=lineWidth)
ax[1, i].set_ylim(minY + 0.05, maxY - 0.05)
title = wildcards[i].replace("=", '')
yloc = plt.MaxNLocator(max_yticks)
if i != 0:
ax[1, i].get_yaxis().set_visible(False)
else:
ax[1, i].set_ylabel('Success', color='black')
ax[1, i].yaxis.set_major_locator(yloc)
#for xValues, s,i in zip(xvalsList, slist, range(0,len(wildcards))):
# plt.plot(xValues,s, paperPlot.sensitivityColorAndPoints, linewidth=lineWidth, color=cm(1.*i/NUM_COLORS))
#plt.plot([min(min(xvalsList[0]), 0), max(xvalsList[0])], [baseline_s, baseline_s], color='k', linestyle='--',
# linewidth=lineWidth)
#plt.ylim((minY, maxY))
#plt.xlabel(wildcard, fontsize=axisFontSize)
#plt.ylabel("Success", fontsize=axisFontSize)
if savefile == '':
plt.show()
else:
plt.savefig(savefile)
def plotStraddelingEdgeOPE(wildcard, savefilename='', format='pdf'):
wu2013GroundTruth = "/Users/Ivan/Files/Data/wu2013"
folderForGraphs = './Visualizations/'
datasetType = 'wu2013'
dataset = Dataset(wu2013GroundTruth, datasetType)
runsNames = glob.glob("./Results/" + wildcard + "*.p")
#runsNames = glob.glob('./Runs/upd=3_hogANDhist_int*.p')
formatSave = 'pdf'
regexp = re.compile("(.*\/)(.+)(.p)")
experimentType = 'default'
d = dict()
for runName in runsNames:
m = re.match(regexp, runName)
name = m.group(2)
run = loadPickle(runName)
results = run.plotMetricsDictEntry['default']
d[name] = results[4:6]
(df1, df2) = reshuffleStraddlingEdgeOPEData(d)
save = False
# Initialize a grid of plots with an Axes for each walk
grid = sns.FacetGrid(df1,
col="$\lambda_e$",
hue="$\lambda_e$",
col_wrap=3,
size=3)
# Draw a horizontal line to show the starting point
grid.map(plt.axhline, y=0.749, ls=":", c=".5")
# Draw a line plot to show the trajectory of each random walk
grid.map(plt.plot, "$\lambda_s$", "Precision", marker="o", ms=4)
grid.set(xticks=np.arange(6) / float(10),
yticks=[0.6, 0.65, 0.7, 0.75, 0.8, 0.85],
xlim=(0, 0.5),
ylim=(0.6, 0.85))
grid.fig.tight_layout(w_pad=1)
# Adjust the tick positions and labels
if savefilename == '':
plt.show()
else:
plt.savefig(savefilename + "precision." + format)
grid = sns.FacetGrid(df2,
col="$\lambda_e$",
hue="$\lambda_e$",
col_wrap=3,
size=3)
grid.map(plt.axhline, y=0.59, ls=":", c=".5")
grid.map(plt.plot, "$\lambda_s$", "Success", marker="o", ms=4)
grid.set(xticks=np.arange(6) / float(10),
yticks=[0.5, 0.55, 0.6, 0.65, 0.7],
xlim=(0, 0.5),
ylim=(0.5, 0.7))
grid.fig.tight_layout(w_pad=1)
if savefilename == '':
plt.show()
else:
plt.savefig(savefilename + "success." + format)
def plotSensitivityUpd(paperPlot, baselineRun, wildcard, savefile=''):
(p,s, xValues) = getPrecisionSuccessGivenWildCard(paperPlot, wildcard)
run = loadPickle(baselineRun)
baseline_p=run.completeMetricDictEntry[0]
baseline_s = run.completeMetricDictEntry[1]
titleFontSize = paperPlot.titleFontSize;
headerFontSize = paperPlot.headerFontSize;
axisFontSize = paperPlot.axisFontSize;
lineWidth = paperPlot.lineWidth;
legendSize = paperPlot.legendSize;
labelsFontSize = paperPlot.labelsFontSize
minY=paperPlot.minY
maxY=paperPlot.maxY
deltaPrecision=paperPlot.deltaPrecision
cm = plt.get_cmap('gist_rainbow')
NUM_COLORS =3
max_yticks = 5
# now everything is sorted
fig, ax = plt.subplots(nrows=1, ncols=2, sharex=True, figsize=(5,4))
fig.subplots_adjust(hspace=5)
#plt.suptitle('Sensitivity analysis for parameter: '+wildcard,fontsize=paperPlot.axisFontSize+4)
i = 0
ax[0].plot(xValues,p, paperPlot.sensitivityColorAndPoints, linewidth=lineWidth, color=cm(1.*i/NUM_COLORS))
ax[0].plot([min(min(xValues),1), max(xValues)], [baseline_p, baseline_p], color='k', linestyle='--', linewidth=lineWidth)
ax[0].set_ylim(minY+0.05+ deltaPrecision,maxY+ deltaPrecision-0.05)
title = wildcard.replace("=",'')
yloc = plt.MaxNLocator(max_yticks)
ax[i].set_ylabel('Precision', color='black')
ax[i].yaxis.set_major_locator(yloc)
#plt.ylim((minY+ deltaPrecision,maxY+ deltaPrecision))
#plt.ylim((minY+ deltaPrecision,maxY+ deltaPrecision))
#ax2 = ax[0,3].twinx()
#ax2.set_ylabel('Precision', color='black')
#plt.xlabel(wildcard, fontsize=axisFontSize)
#plt.ylabel("Precision", fontsize=axisFontSize)
#plt.legend(['filter on','filter off'])
#plt.subplot(2, 4, 5)
ax[1].plot(xValues,s, paperPlot.sensitivityColorAndPoints, linewidth=lineWidth, color=cm(1.*i/NUM_COLORS))
ax[1].plot([min(min(xValues),1), max(xValues)], [baseline_s, baseline_s], color='k', linestyle='--', linewidth=lineWidth)
ax[1].set_ylim(minY+0.05,maxY-0.05)
title = wildcard.replace("=",'')
yloc = plt.MaxNLocator(max_yticks)
i = 0
ax[1].set_ylabel('Success', color='black')
ax[1].yaxis.set_major_locator(yloc)
#for xValues, s,i in zip(xvalsList, slist, range(0,len(wildcards))):
# plt.plot(xValues,s, paperPlot.sensitivityColorAndPoints, linewidth=lineWidth, color=cm(1.*i/NUM_COLORS))
#plt.plot([min(min(xvalsList[0]), 0), max(xvalsList[0])], [baseline_s, baseline_s], color='k', linestyle='--',
# linewidth=lineWidth)
#plt.ylim((minY, maxY))
#plt.xlabel(wildcard, fontsize=axisFontSize)
#plt.ylabel("Success", fontsize=axisFontSize)
if savefile == '':
plt.show()
else:
plt.savefig(savefile)
def plotsensitivity(self, baselineRun,wildcard,savefile=''):
(plotMetricsDict, completeMetricDict) = self.getRuns(wildcard)
run = loadPickle(baselineRun)
baseline_p=run.completeMetricDictEntry[0]
baseline_s = run.completeMetricDictEntry[1]
cm = plt.get_cmap('gist_rainbow')
titleFontSize = self.titleFontSize;
headerFontSize = self.headerFontSize;
axisFontSize = self.axisFontSize;
lineWidth = self.lineWidth;
legendSize = self.legendSize;
labelsFontSize = self.labelsFontSize
minY=self.minY
maxY=self.maxY
deltaPrecision=self.deltaPrecision
# this should be just regular plot: value vs b
# get values from names
runNames = plotMetricsDict.keys()
regExp=re.compile("([\D|=]+)(\d+)")
xValues=list()
for r in runNames:
m=regExp.match(r)
xValues.append((float)(m.group(2)))
# sort xValues
idx_sorted = [i[0] for i in sorted(enumerate(xValues), key=lambda x: x[1])]
xValues=sorted(xValues)
correctNames = [runNames[x] for x in idx_sorted]
# now everything is sorted
plt.figure(figsize=(13,9))
p=list()
s=list()
for name in correctNames:
p.append(completeMetricDict[name][0])
s.append(completeMetricDict[name][1])
plt.suptitle('Sensitivity analysis for parameter: '+wildcard,fontsize=self.axisFontSize+4)
plt.subplot(1,2,1)
plt.plot(xValues,p, self.sensitivityColorAndPoints,linewidth=lineWidth)
plt.plot([min(min(xValues),0), max(xValues)], [baseline_p, baseline_p], color='k', linestyle='--', linewidth=lineWidth)
plt.ylim((minY+ deltaPrecision,maxY+ deltaPrecision))
#plt.title("Precision")
plt.xlabel(wildcard, fontsize=axisFontSize)
plt.ylabel("Precision", fontsize=axisFontSize)
plt.legend(['filter on','filter off'])
plt.subplot(1, 2, 2)
plt.plot(xValues,s, self.sensitivityColorAndPoints, linewidth=lineWidth)
plt.plot([min(min(xValues), 0), max(xValues)], [baseline_s, baseline_s], color='k', linestyle='--',
linewidth=lineWidth)
plt.ylim((minY, maxY))
#plt.title("Success")
plt.xlabel(wildcard, fontsize=axisFontSize)
plt.ylabel("Success", fontsize=axisFontSize)
if savefile == '':
plt.show()
else:
plt.savefig(savefile)
for v in d:
z=VideoResult(v)
videos.append(z)
self.videos=videos
def process(self,cpus=3):
for v in self.videos:
v.process()
if __name__ == "__main__":
wu2013GroundTruth = "/Users/Ivan/Files/Data/Tracking_benchmark"
datasetType = 'wu2013'
# dataset = Dataset(wu2013GroundTruth, datasetType)
# d = dataset.dictData
#
# wholeDataset=DatasetResult(d)
# wholeDataset.process()
#
saveName='objectness_gt_vs_else_pickle.p'
# savePickle(wholeDataset,saveName)
wholeDataset=loadPickle(saveName)
# This experiment shows how does objecteness differs as a function of distance from the object center
