def respond_to_demonstration_letter(self,
demonstration,
letter,
mode='midway',
mode_param=0.5):
demo_stroke = Stroke()
demo_stroke.stroke_from_xxyy(
np.reshape(demonstration, len(demonstration)))
demo_stroke.normalize_wrt_max()
if mode == 'midway':
learned_stroke = stroke.midway(demo_stroke,
self.generated_letters[letter],
mode_param)
self.generated_letters[letter] = learned_stroke
save_learned_allograph(self.child_path, letter, learned_stroke)
_, score = stroke.euclidian_distance(demo_stroke,
self.refs[letter])
if mode == 'simple':
learned_stroke = stroke.weigthedSum(demo_stroke,
self.generated_letters[letter],
mode_param)
self.generated_letters[letter] = learned_stroke
save_learned_allograph(self.child_path, letter, learned_stroke)
_, score = stroke.euclidian_distance(demo_stroke,
self.refs[letter])
return self.shape_message(letter), score
def respond_to_demonstration_word(self, demonstrations, mode='midway'): #mutual_modeling will act here
if mode == 'midway':
for letter,stroke in demonstrations:
learned_stroke = stroke.midway(stroke, self.generated_letters[letter])
self.generated_letters[letter] = learned_stroke
save_learned_allograph(self.robot_data, letter, learned_stroke)
score = stroke.euclidian_distance(demo_stroke, self.refs[letter])
def respond_to_demonstration_letter(self, demonstration, letter, grade, mode='midway'):
demo_stroke = Stroke()
demo_stroke.stroke_from_xxyy(np.reshape(demonstration,len(demonstration)))
#demo_stroke.uniformize()
demo_stroke.normalize_wrt_max()
if mode == 'midway':
learned_stroke = stroke.midway(demo_stroke, self.generated_letters[letter], grade)
self.generated_letters[letter] = learned_stroke
save_learned_allograph(self.robot_path, letter, learned_stroke)
_,score = stroke.euclidian_distance(demo_stroke, self.refs[letter])
#if mode = 'PCA'
#if mode = 'sigNorm' (mixture of sigma-log-normal)
#if mode = 'CNN' (1-D convolutionnal neural networks)
return self.shape_message(letter),score
time = bug_variables[0]
name = bug_variables[1]
ischouchou = bug_variables[2]
if found_demo_bug2:
string2 = str(found_demo_bug2.group("path2"))
string = string_bug + string2
try:
path = np.array(
literal_eval(string)) #.replace('"','')))
graph = Stroke()
graph.stroke_from_xxyy(path)
graph.downsampleShape(70)
if name in prev_strokes:
_, score = stroke.euclidian_distance(
prev_strokes[name], graph)
prev_strokes[name].reset()
prev_strokes[name].stroke_from_xxyy(path)
prev_strokes[name].downsampleShape(70)
else:
score = -1.
prev_strokes[name] = Stroke()
prev_strokes[name].stroke_from_xxyy(path)
prev_strokes[name].downsampleShape(70)
liste = []
liste.append(time)
liste.append(child)
liste.append(name)
liste.append(score)
liste.append(ischouchou)
found_demo = new_demo.search(line)
found_button = new_button.search(line)
found_word = new_word.search(line)
found_repetition = new_repetition.search(line)
found_demo_bug = new_demo_bug.search(line)
found_demo_bug2 = new_demo_bug2.search(line)
found_learn = new_learn.search(line)
if found_repetition:
word_done = True
current_letters = {}
for liste in lines:
letter = liste[2]
if letter in current_demo and letter in current_learn:
_, score = stroke.euclidian_distance(
current_demo[letter],
current_learn[letter])
wr.writerow([
liste[0], liste[1], liste[2], liste[3],
float('nan'),
float('nan'), 1, f2f
])
# lines = []
# current_demo = {}
# current_learn = {}
# feedback = float('nan')
# word_done = False
# nb_fb=1.
if found_learn:
string = str(found_learn.group("path")) #[:-1]
if bug_on_last_line:
time = bug_variables[0]
name = bug_variables[1]
ischouchou = bug_variables[2]
if found_demo_bug2:
string2 = str(found_demo_bug2.group("path2"))
string = string_bug + string2
try:
path = np.array(literal_eval(string)) # .replace('"','')))
graph = Stroke()
graph.stroke_from_xxyy(path)
graph.downsampleShape(70)
if name in prev_strokes:
_, score = stroke.euclidian_distance(prev_strokes[name], graph)
prev_strokes[name].reset()
prev_strokes[name].stroke_from_xxyy(path)
prev_strokes[name].downsampleShape(70)
else:
score = -1.0
prev_strokes[name] = Stroke()
prev_strokes[name].stroke_from_xxyy(path)
prev_strokes[name].downsampleShape(70)
liste = []
liste.append(time)
liste.append(child)
liste.append(name)
liste.append(score)
liste.append(ischouchou)
path = np.array(literal_eval(string))#.replace('"','')))
graph = Stroke()
graph.stroke_from_xxyy(path)
graph.downsampleShape(70)
current_learn[name]=graph
if found_word:
word = str(found_word.group("word"))[:-1]
if word_done:
for liste in lines:
letter = liste[2]
if letter in current_demo and letter in current_learn:
_,score = stroke.euclidian_distance(current_demo[letter], current_learn[letter])
wr.writerow([liste[0],liste[1],liste[2],liste[3],feedback/nb_fb,1,f2f,str(current_learn[letter].x),str(current_learn[letter].y)])
lines = []
current_demo = {}
current_learn = {}
feedback = float('nan')
word_done = False
nb_fb=1.
if found_button:
if str(found_button.group("button"))=="+":
if math.isnan(feedback):
feedback = 1
else: