def main(_):
dataset = Dataset_Youtube8M(dataset_dir=FLAGS.data_dir,
normalization=False,
using_existing_features=False,
flag=FLAGS)
K.clear_session()
learner = LearnerLSTMReg(dataset=dataset,
learner_name='LSTMReg',
flag=FLAGS)
# evaluator = Evaluator_AVEC2016()
learner.learn()
truth, prediction = learner.predict()
predict_yaml_test,
delimiter=",",
fmt="%.3f")
np.savetxt("predict_yaml_train_classes.csv",
predict_yaml_train_classes,
delimiter=",",
fmt="%d")
np.savetxt("predict_yaml_test_classes.csv",
predict_yaml_test_classes,
delimiter=",",
fmt="%d")
return history
# This is also added for candle compliance so that the program can
# still be executed independently from the command line.
def main():
gParameters = initialize_parameters()
run(gParameters)
if __name__ == '__main__':
main()
try:
ke.clear_session()
except AttributeError: # theano does not have this function
pass
train, labels = make_moons(n_samples=1000, noise=0.4, random_state=1)
train_x, test, label_x, label = train_test_split(train,
labels,
test_size=0.3,
random_state=101)
train_x.shape
test.shape
plt.scatter(train_x[:, 0], train_x[:, 1], c=label_x.flatten(), edgecolors='b')
plt.show()
x_min
plt.contourf(train_x[:, 0], train[:, 1], c=label_x)
plt.show()
k.clear_session()
model = Sequential()
model.add(l.Dense(10, activation=l.activations.relu, input_dim=2))
model.add(l.Dropout(.2))
model.add(l.Dense(8, activation=l.activations.tanh))
model.add(l.Dropout(0.2))
model.add(l.Dense(10))
model.add(l.Dropout(.2))
model.add(l.Activation('relu'))
model.add(l.Dense(1, activation=l.activations.sigmoid))
model.compile(optimizer='rmsprop',
loss='binary_crossentropy',
metrics=['accuracy'])
model.summary()
model.fit(x=train_x, y=label_x, batch_size=64, epochs=1000)
model.evaluate(train_x, label_x)
X_train = np.reshape(X_train, (X_train.shape[0], X_train.shape[1], 1))
X_test = np.reshape(X_test, (X_test.shape[0], X_test.shape[1], 1))
model = Sequential()
model.add(LSTM(50, input_shape=(X_train.shape[1], X_train.shape[2])))
model.add(Dense(1))
model.compile(loss='mae', optimizer='adam')
# fit network
history = model.fit(X_train, y_train, epochs=200, batch_size=50, validation_data=(X_test, y_test), verbose=0, shuffle=False)
testPredict = model.predict(X_test)
pred_vec = np.append(pred_vec,testPredict)
K.clear_session()
# plot history
true_vec = y[1:1384,]
def rmse(pred_vec, true_vec):
return np.sqrt(((pred_vec - true_vec) ** 2).mean())
rmse_val = rmse(np.array(pred_vec), np.array(true_vec))
print("rms error is: " + str(rmse_val))
# visually inspect results (requires matplotlib)
#rpca.plot_fit()
import numpy as np
import keras
from keras.models import load_model
from keras.models import Sequential
from keras.layers import Dense, Dropout, Flatten
from keras.layers import Conv2D
from keras.optimizers import Adam
from keras.layers import MaxPooling2D
RCP_dict = {0: "Rock", 1: "Paper", 2: "Scissors"}
cap = cv2.VideoCapture(0)
model = load_model('RockPaperScissors.h5')
while True:
keras.clear_session()
ret, frame = cap.read()
if not ret:
break
#frame=frame.resize(1,300,200,3)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
cropped_img = np.expand_dims(cv2.resize(gray, (200, 300)), -1)
rr = gray.resize(300, 200, 3)
prediction = model.predict(rr, steps=1)
#maxindex = np.argmax(prediction)
print(prediction)
#cv2.putText(frame, prediction, (40+20, 120-60), 1, (255, 255, 255), 2, cv2.LINE_AA)
objects = (['Rock', 'Paper', 'Scissors'])
index = np.arange(len(objects))
