VGG_feature_speconly.py

import numpy as np
import os
from keras.models import Sequential
from keras.layers.convolutional import Convolution2D, MaxPooling2D, ZeroPadding2D
from keras.layers.core import Flatten, Dense, Dropout
from keras.applications.vgg19 import VGG19
from args import parser

args = parser.parse_args()
path_save_VGG= args.VGG_features_path

path_audio_img= args.audio_images_folder_path

path_save_rgb= os.path.join(path_audio_img, 'RGB_sound_VGG.npy')

RGB_sound=np.load(path_save_rgb)
RGB_sound[:,:,:,1]=RGB_sound[:,:,:,0]
RGB_sound[:,:,:,2]=RGB_sound[:,:,:,0]

print('shape RGB sound',RGB_sound.shape)


model = VGG19(weights='imagenet', include_top=False)


preds = model.predict(RGB_sound)


print('pred_reduc taille',preds.shape)

modelbis = Sequential()

modelbis.add(ZeroPadding2D((1,1),input_shape=(7,7,512)))
modelbis.add(MaxPooling2D((2,2), strides=(2,2)))


pred_reduc_pool=modelbis.predict(preds)

print('pool pred_reduc taille',pred_reduc_pool.shape)

pred_reduc_pool = pred_reduc_pool.reshape((-1,512 * 4 * 4))

print('concatenate in size',pred_reduc_pool.shape)


path_save_features= os.path.join(path_save_VGG, 'VGG_feat_speconly')


np.save(path_save_features,pred_reduc_pool)
print('ended successfully')