def get_data_prms(folderName='nicks-house', subFolderName='Angle1Lighting1'): dArgs = edict() dArgs.folderName = folderName dArgs.subFolderName = subFolderName #Save the parameters in a database dbFile = DB % 'folder-data' dArgs.prmStr = msq.get_sql_id(dbFile, dArgs) return dArgs
def get_vatic_prms(**kwargs):
dArgs = edict()
pths = get_basic_paths()
#Directory where vatic data is
dArgs.vaticDr = pths.vatic.dr
#Directory where vatic data should be output
dArgs.vaticOpDr = '/mnt/HardDrive/common/vatic'
#merge_threshold - dictates how vatic annotatiosn are merged
#no need to change this
dArgs.mergeThreshold = 0.5
#Save the parameters in a database
dbFile = DB % 'vatic'
dArgs.prmStr = msq.get_sql_id(dbFile, dArgs,
ignoreKeys=['vaticDr', 'vaticOpDr'])
return dArgs
def set_rcnn_prms(**kwargs):
dArgs = edict()
#Object class that needs to be detected
dArgs.targetClass = kwargs.get('targetClass', 'person')
#NMS
dArgs.nmsThresh = kwargs.get('nmsThresh', 0.3)
#Detection Confidence
dArgs.confThresh = kwargs.get('confThresh', 0.8)
dArgs.topK = kwargs.get('topK', 5)
#What classnames was the detector trained on.
dArgs.trainDataSet = kwargs.get('trainDataSet', 'coco')
#The net to be used
dArgs.netName = kwargs.get('netName', 'vgg16-coco-rcnn')
dArgs = cu.get_defaults(kwargs, dArgs, True)
#Save the parameters in a database
dbFile = DB % 'rcnn'
dArgs._mysqlid = msq.get_sql_id(dbFile, dArgs)
#verify that the target class is detectable by the model
allCls = dataset2classnames(dArgs.trainDataSet)
assert dArgs.targetClass in allCls
return dArgs
