def outfacing(lock, receiver, ttsque=None, finder=finder):
cv2.namedWindow('visuals')
processor = faceprocessor()
extractor = modelInferance()
processor.setSource(cv2.VideoCapture(0))
processor.setFinder(finder().nextFaces)
faceDict = FaceDictionary()
datawindow = window()
dnn = model()
dnn.load_previous()
generator = datagenerator()
clf, revmap = load_classifier(lock)
cacher = feature_cacher()
kill = False
cacheFlag = False
label = None
cv2.namedWindow('visual')
ix = 0
while cv2.waitKey(10) < 0 and not kill:
temp = processor.getFaces()
if receiver and receiver.poll(0.001):
signal = receiver.recv()
print(signal)
if signal.find('label:') >= 0:
label = signal[signal.find(':') + 1:]
cacheFlag = True
if signal == 'STOPCACHE':
cacheFlag = False
if signal == 'KILL':
kill = True
if signal == 'Done Thinking':
clf, revmap = load_classifier(lock)
print('HOT SWAPPED CLASSIFIER')
if temp is None:
continue
frame, fbbs, wfaces = temp
feats = extractor.getFeatures(wfaces)
#faceDict.update(fbbs,feats)
#index,fbbs,feats = faceDict.getobjs()
index = np.ones(len(fbbs))
clss, score, tmap = handledata(dnn, cacher, generator, cacheFlag,
label, feats, wfaces, datawindow)
if tmap is not None:
revmap = tmap
if score is not None:
isUnknown = score < 8.0
else:
isUnknown = False
sendmesg(isUnknown, 'What are you doing here? Relinquish thine name. ',
ttsque)
ix += 1
visualize(frame, fbbs, wfaces, clss, revmap, isUnknown, index)
# print(clss)
extractor.kill()
return 1
def ROS_Facing(lock):
processor = faceprocessor()
extractor = modelInferance()
faceDict = FaceDictionary()
datawindow = IdWindows()
dnn = model()
dnn.load_previous()
generator = datagenerator()
clf, revmap = load_classifier(lock)
cacher = feature_cacher()
opts = {'kill': False, 'cacheFlag': False, 'label': None}
def destroy():
extractor.kill()
def callback(data, options={}):
bbs, img = data
temp = processor.get_cb_faces(img, bbs)
if temp == None:
return
frame, fbbs, wfaces = temp
feats = extractor.getFeatures(wfaces)
faceDict.update(fbbs, feats)
index, fbbs, feats = faceDict.getobjs()
if 'label' in options and options['label'] is not None:
opts['label'] = options['label']
if 'store' in options:
opts['cacheFlag'] = options['store']
clss, score, tmap = handledata(dnn, cacher, generator,
opts['cacheFlag'], opts['label'], feats,
wfaces, datawindow, index)
if 'save' in options and options['save'] is not None:
options['save'] = None
dnn.save(filename=options['save'])
if 'load' in options and options['load'] is not None:
options['load'] = None
dnn.load_previous(filename=options['load'])
clasmap = ''
if tmap is not None:
clasmap = tmap
else:
clasmap = revmap
isUnknown = {
k: v
for k, v in zip(index, np.zeros((1, len(index)), dtype=bool))
}
if len(isUnknown.keys()) == 0:
isUnknown = [False]
for id in index:
if id in score:
isUnknown[id] = score[id] < 8.0
else:
isUnknown[id] = False
return visualize(frame, fbbs, wfaces, clss, clasmap, isUnknown, index)
return callback, destroy
# coding: utf-8 import numpy as np from tensorflowclassifier import model mymodel = model() from datagen import datagenerator dg = datagenerator() mymodel.load_previous mymodel.load_previous() dg.load_sectioned_features() mymodel.input_data(dg.testingfeats[0], dg.testinglbls[0]) dg.testingfeats mymodel.input_data(dg.testingfeats[0], dg.testinglbls[0]) mymodel.input_data(dg.trainingfeats[0], dg.traininglbls[0]) dg.traininglbls[0] mymodel.input_data(dg.trainingfeats[2], dg.traininglbls[2]) dg.traininglbls[2] p = mymodel.pred() p np.sum(p.argmax() == 2) / p.shape[0] p p.argmax() p.argmax(axis=1) mymodel.input_data(dg.trainingfeats[3], dg.traininglbls[3]) p = mymodel.pred() p p.argmax(axis=1) p.argmax(axis=1) == 3 np.sum(p.argmax(axis=1) == 3) / p.shape[0] p.argmax(axis=1) p.argsort()[-2:]