def train_classifier(genres=[
'Chases', 'Dance', 'Eating', 'Fight', 'Heated_Discussions',
'Normal_Chatting', 'Romance', 'Running', 'Tragic'
],
model_name=default_model_name):
trainingData = []
trainingLabels = []
num_of_classes = len(genres)
print("Number of classes:", num_of_classes)
for genreIndex, genre in enumerate(genres):
try:
genreFeatures = load_pkl(genre + "_ultimate_OF_" +
default_model_name)
genreFeatures = np.array([np.array(f)
for f in genreFeatures]) # numpy hack
except Exception as e:
print(e)
return
print("OK.")
for videoFeatures in genreFeatures:
randomIndices = range(len(videoFeatures))
selectedFeatures = np.array(videoFeatures[randomIndices])
for feature in selectedFeatures:
trainingData.append(feature)
trainingLabels.append([genreIndex])
trainingData = np.array(trainingData)
trainingLabels = np.array(trainingLabels)
print("Train data shape : ", trainingData.shape)
print(trainingLabels.shape)
model = spatial_model(trainingData)
model.compile(optimizer='sgd',
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
batch_size = 32
nb_epoch = 100
history = model.fit(trainingData,
trainingLabels,
batch_size=batch_size,
nb_epoch=nb_epoch)
modelOutPath = 'data/models/temporal' + model_name + '_' + str(
num_of_classes) + "g_bs" + str(batch_size) + "_ep" + str(
nb_epoch) + ".h5"
model.save(modelOutPath)
print("Model saved at", modelOutPath)
plt.plot(history.history["acc"])
plt.title("model accuracy")
plt.ylabel("accuracy")
plt.xlabel("epoch")
plt.show()
plt.plot(history.history["loss"])
plt.title("model loss")
plt.ylabel("loss")
plt.xlabel("epoch")
plt.show()
def train_classifier(genres=['romance', 'horror', 'action'],
num_of_videos=100):
"""Gather data for selected genres"""
trainingData = []
trainingLabels = []
num_of_random_frames = 75
num_of_classes = len(genres)
print "Number of classes:", num_of_classes
for genreIndex, genre in enumerate(genres):
print "Looking for pickle file: data/{0}{1}.p".format(
genre, str(num_of_videos)),
try:
genreFeatures = load_pkl(genre + str(num_of_videos))
genreFeatures = np.array([np.array(f)
for f in genreFeatures]) # numpy hack
except Exception as e:
print e
return
print "OK."
for videoFeatures in genreFeatures:
if len(videoFeatures) > num_of_random_frames:
randomIndices = np.random.randint(0, len(videoFeatures),
num_of_random_frames)
selectedFeatures = np.array(videoFeatures[randomIndices])
for feature in selectedFeatures:
trainingData.append(feature)
trainingLabels.append([genreIndex])
trainingData = np.array(trainingData)
trainingLabels = np.array(trainingLabels)
print trainingData.shape
print trainingLabels.shape
# trainingLabels = to_categorical(trainingLabels, num_of_classes)
print trainingLabels
# trainingLabels = trainingLabels.reshape((-1,num_of_classes))
"""Initialize the mode"""
model = spatial_model(num_of_classes)
model.compile(optimizer='sgd',
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
"""Start training"""
batch_size = 32
nb_epoch = 50
model.fit(trainingData,
trainingLabels,
batch_size=batch_size,
nb_epoch=nb_epoch) #, callbacks=[remote])
modelOutPath = 'data/models/spatial_' + str(num_of_classes) + "g_bs" + str(
batch_size) + "_ep" + str(nb_epoch) + "_nf_" + str(
num_of_random_frames) + ".h5"
model.save(modelOutPath)
print "Model saved at", modelOutPath
def train_classifier(genres=['comedy', 'horror', 'action'],
model_train='spatial',
model_name=default_model_name,
epoch=100):
"""Gather data for selected genres"""
trainingData = []
trainingLabels = []
# num_of_random_frames = 75
num_of_classes = len(genres)
print("Number of classes:", num_of_classes)
for genreIndex, genre in enumerate(genres):
# print "Looking for pickle file: data/{0}{1}.p".format(genre, str(num_of_videos)),
try:
print("train/" + genre + "_train_" + model_name)
genreFeatures = load_pkl("train/" + genre + "_train_" + model_name)
genreFeatures = np.array([np.array(f)
for f in genreFeatures]) # numpy hack
except Exception as e:
print(e)
return
print("OK.")
for videoFeatures in genreFeatures:
"""to get all frames from a video -- hacky"""
randomIndices = range(len(videoFeatures))
selectedFeatures = np.array(videoFeatures[randomIndices])
for feature in selectedFeatures:
trainingData.append(feature)
trainingLabels.append([genreIndex])
trainingData = np.array(trainingData)
trainingLabels = np.array(trainingLabels)
print(trainingData.shape)
print(trainingLabels.shape)
# trainingLabels = to_categorical(trainingLabels, num_of_classes)
print(trainingLabels)
# trainingLabels = trainingLabels.reshape((-1,num_of_classes))
"""Initialize the mode"""
if (model_train == 'mlp'):
model = mlp_model(num_of_classes, trainingData.shape[1])
model.compile(optimizer='adadelta',
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
elif (model_train == 'lstm'):
num_of_frames = 9
num_of_samples = len(trainingLabels)
trainingDataTensor = np.zeros(
(num_of_samples, num_of_frames, trainingData.shape[1]))
print(len(trainingData))
for sampleIndex in range(num_of_samples):
for vectorIndex in range(num_of_frames):
try:
trainingDataTensor[sampleIndex][
vectorIndex] = trainingData[sampleIndex][vectorIndex]
except Exception as e:
print(e)
continue
print(trainingDataTensor.shape)
trainingData = trainingDataTensor
model = lstm_model(num_of_classes, input_dim=trainingData.shape[2])
model.compile(optimizer='sgd',
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
elif (model_train == 'spatial'):
model = spatial_model(num_of_classes, trainingData.shape[1])
model.compile(optimizer='sgd',
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
"""Start training"""
batch_size = 32
#b_epoch = 100
model.fit(trainingData,
trainingLabels,
batch_size=batch_size,
epochs=int(epoch)) #, callbacks=[remote])
modelOutPath = 'data/models/' + model_train + '_' + model_name + '_' + str(
num_of_classes) + "g_bs" + str(batch_size) + "_ep" + str(epoch) + ".h5"
model.save(modelOutPath)
print("Model saved at", modelOutPath)