frames_folder = 'path/to/folder'
X_data = []
y_data = []
list_of_sent = os.listdir(frames_folder)
print (list_of_sent)
class_num = 0
time_steps = 0
frames = []
for i in list_of_sent:
    classes_folder = str(frames_folder + '/' + i) #path to each class
    print (classes_folder)
    list_of_frames = os.listdir(classes_folder)
    time_steps= 0
    frames = []
    for filename in  sorted(list_of_frames):
        if ( time_steps == total_timesteps ):
            X_data.append(frames) #appending each tensor of 8 frames resized to 110,110
            y_data.append(class_num) #appending a class label to the set of 8 frames
            time_steps = 0
            frames = []
        else:
            time_steps+=1
            filename = cv2.imread(vid + '/' + filename)
            filename = cv2.resize(filename,(width, height),interpolation=cv2.INTER_AREA)
            frames.append(filename)


    class_num+=1
X_data = np.array(X_data)
y_data = np.array(y_data)

print (X_data.shape)
X_data = X_data.reshape(samples ,timesteps,width* height*channels)

#Encoding Targets with OneHotEncoding
y_data = y_data.reshape(len(y_data), 1)
encoder = OneHotEncoder(sparse=False)
y_encoded = encoder.fit_transform(y_data)

#Splitting in Train and Test Data
X_train, X_test, y_train, y_test = train_test_split(X_data, y_encoded, test_size=0.2, shuffle=True, random_state=100 , stratify=y_encoded)

model = Sequential()
model.add(LSTM(16, return_sequences=False, input_shape=(timesteps, width* height*channels)))
model.add(Dense(classes, activation='softmax'))
model.summary()

adam = optimizers.Adam(lr=0.00001)#, decay=0.000001)
model.compile(optimizer = adam, loss='categorical_crossentropy', metrics=['accuracy'])#, sample_weight_mode='temporal')

history = model.fit(X_train, y_train, batch_size=16, epochs = 48, validation_data=(X_test, y_test))