def auto_drive(img_name):
res = 1 if not detect else check_position()
if res == 1:
# If we are in the right track
if len(sa_lst) == len(block_lst):
block_lst.append([])
md_img, _ = process_image(img_name, None, False)
pred_act = model.predict(np.array([md_img]))[0]
print("Lft: %.2f | Fwd: %.2f | Rght: %.2f" %
(pred_act[1], pred_act[0], pred_act[2]))
act_i = np.argmax(pred_act)
if block_lst:
while pred_act[act_i] >= 0 and act_i in block_lst[-1]:
pred_act[act_i] = -1.
act_i = np.argmax(pred_act)
if act_i == -1:
block_lst.pop()
reverse_motion()
else:
send_control(act_i, img_name)
return pred_act, act_i
elif res == -1:
# If we cannot detect where we are
print("Error: cannot identify position")
return -1, -1
else:
# If we are outside
try:
reverse_motion()
except:
print("Error: cannot reverse an action")
def draw_predict(): ## Get prediction steering angle md_img, _ = process_image(img_name, None, False) pred_prob = model.predict(np.array([md_img]))[0] pred_angle = get_steering(np.argmax(pred_prob)) ## Draw green prediction line x_shift = r * sin(pred_angle) y_shift = r * cos(pred_angle) cv2.line(img,(int(width / 2 - x_shift), int(height - y_shift)),(width // 2, height),(0,255,0),2) return pred_angle, pred_prob
def predict_animal(file):
print("Predicting .................................")
ar = convert_to_array(file)
ar = ar / 255
a = [ar]
a = np.array(a)
score = model.predict(a, verbose=1)
label_index = np.argmax(score)
acc = np.max(score)
animal = get_animal_name(label_index)
print("The predicted animal is a " + animal + " with accuracy = " + str(acc))
def draw_predict(): ## Get prediction steering angle md_img, _ = process_image(img_name, None, False) pred_angle = model.predict(np.array([md_img]))[0][0] if pred_angle > max_angle: pred_angle = max_angle elif pred_angle < -max_angle: pred_angle = -max_angle ## Draw green prediction line x_shift = 1.2 * r * sin(pred_angle) y_shift = 1.2 * r * cos(pred_angle) cv2.line(img,(int(width / 2 - x_shift), int(height - y_shift)),(width // 2, height),(0,255,0),3) return pred_angle
def auto_drive(img_name):
md_img, _ = process_image(img_name, None, False)
pred_angle = model.predict(np.array([md_img]))[0][0]
if pred_angle >= max_angle / 2.0:
act_i = 1
if pred_angle > max_angle:
pred_angle = max_angle
elif pred_angle <= -max_angle / 2.0:
act_i = 2
if pred_angle < -max_angle:
pred_angle = -max_angle
else:
act_i = 0
# send_control(act_i)
return pred_angle, act_i
def telemetry(sid, data):
# The current image from the center camera of the car
img_str = data["image"]
speed = float(data["speed"])
# Set the throttle.
throttle = 1.2 - (speed / target_speed)
# read and process image
image_bytes = BytesIO(base64.b64decode(img_str))
image, _ = process_image(image_bytes, None, False)
# make prediction on steering
sa = model.predict(np.array([image]))[0][0]
print(sa, throttle)
send_control(sa, throttle)
def label_predict(test_data):
okt = Okt()
clean_test = []
for x in tqdm(test_data):
if type(x) == str:
clean_test.append(
train.preprocessing(x,
okt,
remove_stopwords=True,
stop_words=stop_words))
else:
clean_test.append([])
test_tokenizer = Tokenizer()
test_tokenizer.fit_on_texts(clean_train_text)
test_sequences = test_tokenizer.texts_to_sequences(clean_test)
test_word_vocab = test_tokenizer.word_index
MAX_SEQUENCE_LENGTH = 30
test_inputs = pad_sequences(test_sequences,
maxlen=MAX_SEQUENCE_LENGTH,
padding='post')
x_test = train.vectorize_sequences(test_inputs)
predictions = model.predict(x_test)
test_predict = []
for i in range(len(test_data)):
a = np.argmax(predictions[i])
test_predict.append(a)
PREDICT = []
for i in range(len(test_data)):
for j in range(len(LABEL)):
if test_predict[i] == LABEL[j][0]:
PREDICT.append(LABEL[j][1])
with open('result.txt', 'w') as f:
for line in PREDICT:
f.write(line)
f.write("\n")
def loop(): action=np.zeros(5) while self.control: #action = (self.cost_position, action, maxiter=1000, maxfun=10000) #action = fmin_bfgs(f=self.cost_position, x0=action, gtol=0.1, epsilon=0.001) #action = fmin_ncg(f=self.cost_position, fprime=None, x0=action, epsilon=0.001) (action, nfeval, rc) = fmin_tnc(self.cost_position, approx_grad=True, fprime=None, x0=action, epsilon=0.001) print action #exit() pred_y = model.predict(np.hstack([latest, action])) action[4] = 0 action[0] = 0 ''' if len(np.zeros(action)) == 5: print "no data for command, doing a random" action = np.random.rand(5) - 0.5 action[4] = 0 action[0] = 0 ''' x = np.hstack([latest, action]) y = applyMove(action) print " self.target ", self.target print " x ", x print " pred dx ", pred_y print " actual ", y model.update(x,y)
plt.ylabel('dice_coef')
plt.xlabel('Epoch')
plt.legend(['Train', 'Test'], loc='upper left')
plt.show()
# Draw Loss
plt.plot(results.history['loss'])
plt.plot(results.history['val_loss'])
plt.title('Model loss')
plt.ylabel('Loss')
plt.xlabel('Epoch')
plt.legend(['Train', 'Test'], loc='upper left')
plt.show()
# Getting the predictions
predictions = model.predict([X_scan, X_seg])
predictions = predictions[:, :, :, 0] # ditch the channel dimension
print(predictions.shape)
# Writing the results to the output file
for i in range(predictions.shape[0]):
output_file = "./output/output" + str(i + 1)
with open(output_file, 'w+') as f:
matrix_to_write = predictions[i]
np.savetxt(output_file, matrix_to_write)
# Visualize the segmentations
fig, axs = plt.subplots(1, 4, figsize=(15, 15))
for i in range(predictions.shape[0]):
seg_to_show = predictions[i]
# from lib code
model.layers[2].load_weights('weights/weights.h5')
anchor_audios = glob.glob('data/{}/*.wav'.format(
sorted(os.listdir('data'))[0]))
anchor_audio = audio2vector(anchor_audios[random.randint(
0,
len(anchor_audios) - 1)])
threshold = 0.3
with open('weights/threshold.txt') as f:
threshold = float(f.read())
text = 0
try:
write('sample.wav', fs, myrecording) # Save as WAV file
audio = audio2vector('sample.wav')
prediction = model.predict([[audio.reshape((20, 400, 1))],
[anchor_audio.reshape((20, 400, 1))]])
prediction = prediction[0, 0]
text = 1 if prediction < threshold else 0
# print(threshold, prediction)
except Exception as e:
pass
print(text, end='')
def apply_mask(img, mask_size, step, nums):
global model_time
width = 320
height = 120
mask = img.copy()
map0 = Image.new("RGB", (width, height))
map1 = Image.new("RGB", (width, height))
map2 = Image.new("RGB", (width, height))
map3 = Image.new("RGB", (width, height))
pixels = mask.load()
pixels0 = map0.load()
pixels1 = map1.load()
pixels2 = map2.load()
pixels3 = map3.load()
for i in range(mask_size):
for j in range(mask_size):
if i < height and j < width:
pixels[j, i] = (127, 127, 127)
for i in range(0, height, step):
for j in range(0, width, step):
new_img = img.copy()
pixels = new_img.load()
for r in range(i - mask_size // 2, i + (mask_size + 1) // 2):
for c in range(j - mask_size // 2, j + (mask_size + 1) // 2):
if r >= 0 and r < height and c >= 0 and c < width:
pixels[c, r] = (127, 127, 127)
model_start = datetime.datetime.now()
aimage = img_to_array(new_img)
aimage = aimage.astype(np.float32) / 255
aimage = aimage - 0.5
res = model.predict(np.array([aimage]))[0]
model_time += datetime.datetime.now() - model_start
# print(res)
for r in range(i, min(i + step, height)):
for c in range(j, min(j + step, width)):
pixels0[c, r] = (int(
(1 - res[0]) * 255), 0, int(res[0] * 255))
pixels1[c, r] = (int(
(1 - res[1]) * 255), 0, int(res[1] * 255))
pixels2[c, r] = (int(
(1 - res[2]) * 255), 0, int(res[2] * 255))
pixels3[c, r] = (int(
(1 - res[3]) * 255), 0, int(res[3] * 255))
draw0 = ImageDraw.Draw(map0)
draw1 = ImageDraw.Draw(map1)
draw2 = ImageDraw.Draw(map2)
draw3 = ImageDraw.Draw(map3)
draw = ImageDraw.Draw(mask)
draw0.text((0, 0), "Forward %.2f" % nums[0], (0, 255, 0), font)
draw1.text((0, 0), "Left %.2f" % nums[1], (0, 255, 0), font)
draw2.text((0, 0), "Right %.2f" % nums[2], (0, 255, 0), font)
draw3.text((0, 0), "Reverse %.2f" % nums[3], (0, 255, 0), font)
draw.text((0, 0), str(mask_size), (0, 255, 0), font)
final = Image.new("RGB", (width * 2, 12 * height))
img = img.resize((640, 240), PIL.Image.ANTIALIAS)
map0 = map0.resize((640, 240), PIL.Image.ANTIALIAS)
map1 = map1.resize((640, 240), PIL.Image.ANTIALIAS)
map2 = map2.resize((640, 240), PIL.Image.ANTIALIAS)
map3 = map3.resize((640, 240), PIL.Image.ANTIALIAS)
mask = mask.resize((640, 240), PIL.Image.ANTIALIAS)
final.paste(img, (0, 0, 640, 240))
final.paste(map0, (0, 240, 640, 480))
final.paste(map1, (0, 480, 640, 720))
final.paste(map2, (0, 720, 640, 960))
final.paste(map3, (0, 960, 640, 1200))
final.paste(mask, (0, 1200, 640, 1440))
images.append(final)
map2 = map2.resize((640, 240), PIL.Image.ANTIALIAS)
map3 = map3.resize((640, 240), PIL.Image.ANTIALIAS)
mask = mask.resize((640, 240), PIL.Image.ANTIALIAS)
final.paste(img, (0, 0, 640, 240))
final.paste(map0, (0, 240, 640, 480))
final.paste(map1, (0, 480, 640, 720))
final.paste(map2, (0, 720, 640, 960))
final.paste(map3, (0, 960, 640, 1200))
final.paste(mask, (0, 1200, 640, 1440))
images.append(final)
inp = Image.open(sys.argv[1])
model = model(True, (120, 320, 3), tr_model=sys.argv[2])
inp = inp.resize((320, 120))
font = ImageFont.truetype("/System/Library/Fonts/SFNSText.ttf", 16)
aimage = img_to_array(inp)
aimage = aimage.astype(np.float32) / 255
aimage = aimage - 0.5
res = model.predict(np.array([aimage]))[0]
for i in range(1, 120, 4):
apply_mask(inp, i, 5, res)
img = images.pop(0)
img.save("map" + str(i) + ".jpg")
print(datetime.datetime.now() - start)
print(model_time)