pygame.init()
pygame.display.set_mode((w, h), pygame.constants.RESIZABLE | pygame.constants.DOUBLEBUF | pygame.constants.RLEACCEL, 32)
screen = pygame.display.get_surface()
print ("Reading gaze data...")
dw = drawgc_wrapper()
ds.target_fps = 30
ds.total_frame = len(truth_files)
last_time = time.time()
clock = pygame.time.Clock()
# Whether to convert npz saliency maps first
if convert_npz:
print('Converting npz to jpgs..')
# Check if target dir exits
check_dirs.check_dir(dir_save_png_map)
# npz => jpg Conversion
temp_id = ds.cur_frame_id - 1
while temp_id < ds.total_frame:
print('converting frame:', truth_files[temp_id-1])
m = cm.ScalarMappable(cmap='jet')
npz_img = np.load(dir_load_npz_saliency_map + key +
truth_files[temp_id-1].strip('frame').strip('.jpg') + '.npz')[key]
rgb_img = m.to_rgba(npz_img)[:,:,:3]
plt.imsave(dir_save_png_map + truth_files[temp_id-1], rgb_img) # Save jpg
temp_id += 1
# Load saliency maps and play
print('\n loading ground truth and predicted saliency maps..')
while ds.cur_frame_id-1 < ds.total_frame:
print('frame:', truth_files[ds.cur_frame_id-1])
import os
import config
import shutil
import check_dirs
import scipy.io as sio
import numpy as np
import read_frameInfo
matfn = config.groundtruth_file
data = sio.loadmat(matfn)
fix_index = data['fixation_frames']
fix_index = fix_index.astype(np.int32)
subject = 1 # Can be 1,2,3
check_dirs.check_dir(config.dir_to_save_frames_in_use)
frame_sets = [
x for x in os.listdir(config.dir_to_load_frames) if x.endswith('.jpg')
] # Load frames
print('Number of frames:', len(frame_sets))
# Import frame idx
subFrames, _, _ = read_frameInfo.frameInfo('./frameInfo.mat', subject=subject)
for frame in frame_sets:
index = int(frame.strip('frame').strip('.jpg')) # Get loaded frame index
# Range of frames in use
if index in subFrames and fix_index[index] == 1:
shutil.copy(config.dir_to_load_frames + frame,
config.dir_to_save_frames_in_use)
try:
"""
Created on Fri Jun 15 17:41:27 2018
@author: david
"""
'''To use ffmpeg: http://zulko.github.io/blog/2013/09/27/read-and-write-video-frames-in-python-using-ffmpeg/'''
#%%
import cv2
import check_dirs
print('OpenCV version: ', cv2.__version__)
dir_to_load_video = './'
dir_to_save_frames = './frames/'
check_dirs.check_dir(dir_to_save_frames)
pixel = {'x': 1024, 'y': 576} # Target size
cap = cv2.VideoCapture(dir_to_load_video + 'world.mp4') #Obtain video
ret, frame = cap.read() #Capture each frame
count = 0
ret = True
while ret:
frame = cv2.resize(frame, (pixel['x'], pixel['y']), interpolation =cv2.INTER_AREA)
cv2.imwrite(dir_to_save_frames + "frame%d.jpg" % count, frame) # save frames as JPEG files
ret, frame = cap.read()
print ('Read a new frame: ', count)
count += 1
Created on Thu Dec 27 22:07:04 2018
@author: dawei
"""
import os
import check_dirs
import vggish_inference_demo
import tryread
dir_to_save_csv = './csv_files/'
dir_to_save_tf = './tf_records/'
dir_to_load_wav = './wav_files/'
check_dirs.check_dir(dir_to_save_csv)
check_dirs.check_dir(dir_to_save_tf)
if __name__ == '__main__':
# Load wav files
wav_list = [x for x in os.listdir(dir_to_load_wav) if x.endswith('.wav')]
print('Number of wav files:', len(wav_list))
# Run the VGGish model
for wav in wav_list:
wav = wav.strip('.wav')
try:
vggish_object = vggish_inference_demo.vggish(dir_to_load_wav + wav,
dir_to_save_tf + wav)
vggish_object.set_all_flags()
vggish_object.run()
except:
#%%
def reshape(training_data_fit):
# Reshape training data as (#,128,1) for CNN
training_data_fit = np.reshape(
training_data_fit,
(training_data_fit.shape[2], 128, training_data_fit.shape[0]))
training_data_fit = training_data_fit[:, :, :, np.newaxis]
return training_data_fit
#%%
if __name__ == "__main__":
# Create a path to save model
check_dirs.check_dir(save_model_dir)
# Load target classes
classes = pd.read_csv(label_path + 'UCLA_data_codebook.csv').values[:, 0]
classes = classes.tolist()
class_to_select = [e for e in classes if target in e]
print('target class:', class_to_select)
# Load regression labels (values)
UCLA_data = pd.read_csv(label_path + 'UCLA_data.csv')
# Replace empty str as null, then drop
UCLA_data = UCLA_data.replace(' ', np.nan, inplace=False)
UCLA_data = UCLA_data.dropna(how='any')
# File idx, int
csid = UCLA_data.values[:, 0].tolist()
# Labels, float
features_new, labels_new, groups_new = get_contextual(3, features, labels,
groups)
features_1_test, labels_1_test, groups_1_test = get_contextual(
3, features_1, labels_1, groups_1)
features_2_test, labels_2_test, groups_2_test = get_contextual(
3, features_2, labels_2, groups_2)
features_m_test, labels_m_test, groups_m_test = get_contextual(
3, features_mixed, labels_mixed, groups_mixed)
#%%
RF = False
if RF == True:
seed(0)
predict = False # True, False
save_model_path = './models/RF/'
check_dirs.check_dir(save_model_path)
#kfold = StratifiedKFold(n_splits=5, shuffle=False, random_state=None)
lppo = LeavePGroupsOut(n_groups=1)
fold_no = 1
f1_per_fold, acc_per_fold, pre_per_fold, rec_per_fold = [], [], [], []
features = np.reshape(features, (features.shape[0], features.shape[1]))
# training
if not predict:
#for train, test in kfold.split(features, labels):
for train, test in lppo.split(features_new,
labels_new,
groups=groups_new):
feat_train, labels_train = features_new[train], labels_new[
train].reshape((len(labels_new[train])))
feat_test, labels_test = features_new[test], labels_new[
test].reshape((len(labels_new[test])))
# -*- coding: utf-8 -*-
"""
Created on Sat Nov 17 17:54:19 2018
@author: david
"""
import config
import check_dirs
import os
import numpy as np
import cv2
# Check and create necessary dir
dir_groundtruth = config.dir_to_save_groundtruth
check_dirs.check_dir(dir_groundtruth[:-1]+'_flip/')
dir_heatmap = config.dir_to_save_heatmap
check_dirs.check_dir(dir_heatmap[:-1]+'_flip/')
dir_img = config.dir_to_save_frames_in_use
check_dirs.check_dir(dir_img[:-1]+'_flip/')
frame_sets = [x for x in os.listdir(dir_img) if x.endswith('.jpg')] # Load frames
print('Number of frames:', len(frame_sets))
heatmap_sets = [x for x in os.listdir(dir_heatmap) if x.endswith('.npz')] # Load heatmaps
print('Number of heatmaps:', len(heatmap_sets))
groundtruth_sets = [x for x in os.listdir(dir_groundtruth) if x.endswith('.jpg')] # Load groundtruth
print('Number of groundtruth:', len(groundtruth_sets))
"""
import cv2
import pandas as pd
import scipy.io as sio
import os
import numpy as np
import config
import video
import heatmap
import check_dirs
#%%
'''Visualize groundtruth coordinates and generate groundtruth videos/heatmaps'''
if __name__ == '__main__':
check_dirs.check_dir(config.dir_to_save_groundtruth)
check_dirs.check_dir(config.dir_to_save_heatmap)
check_dirs.check_dir(config.dir_write_video)
groundtruth_log = open(config.dir_to_save_groundtruth_log + "/log2.txt",
'w')
'''Import groundtruth coordinates and frame indeces from csv'''
matfn = 'G:/Research2/sem2 w1/Walking_data_Subj1and2/Subj2/por.mat'
data = sio.loadmat(matfn)
norm_pos_x = data['porX'] * 1.0 / 1920
norm_pos_y = data['porY'] * 1.0 / 1080
norm_pos_x = norm_pos_x.T
norm_pos_y = norm_pos_y.T
heatmap_object = heatmap.heatmap(config.pixel['x'], config.pixel['y'])
'''Generate and save groundtruth frames/heatmap/txt'''
frame_sets = [
x for x in os.listdir(config.dir_to_load_frames) if x.endswith('.jpg')
"""
Created on Wed Dec 12 2018
@author: david
"""
import os
import subprocess
import check_dirs
dir_to_save = './wav_files/'
dir_to_load_mp4 = './T1_Audio_Files/'
check_dirs.check_dir(dir_to_save)
mp4_list = [x for x in os.listdir(dir_to_load_mp4)
if x.endswith('.mp4')] # Load mp4 files
print('Number of audio files:', len(mp4_list))
for item in mp4_list:
# Format: 16-bit depth, sampling rate: from source
command = "ffmpeg -i " + dir_to_load_mp4 + item + " -vn -sample_fmt s16 " + dir_to_save + item.strip(
'.mp4') + ".wav"
subprocess.call(command, shell=True)
print('Done')
import metrics
import numpy as np
import config
import os
import itti_model
import heatmap
import check_dirs
#%%
dir_to_load_heatmap = config.dir_to_save_heatmap
dir_to_load_log = config.dir_to_save_log
dir_to_load_frames = config.dir_to_load_frames
y_pos, x_pos = config.pixel['y'], config.pixel['x']
if __name__ == '__main__':
check_dirs.check_dir(config.dir_to_save_log)
check_dirs.check_dir(config.dir_to_save_saliency_map)
check_dirs.check_dir(config.dir_to_save_antisaliency_map)
check_dirs.check_dir(config.dir_to_save_chance_map)
NSS_itti = open(config.dir_to_save_log + "/NSS_itti.txt", 'w')
AUC_itti = open(config.dir_to_save_log + "/AUC_itti.txt", 'w')
CC_itti = open(config.dir_to_save_log + "/CC_itti.txt", 'w')
KL_itti = open(config.dir_to_save_log + "/KL_itti.txt", 'w')
NSS_itti.write('index' + ' ' + 'saliency' + ' ' + 'chance' + ' ' +
'anti_saliency' + '\n')
AUC_itti.write('index' + ' ' + 'saliency' + ' ' + 'chance' + ' ' +
'anti_saliency' + '\n')
CC_itti.write('index' + ' ' + 'saliency' + ' ' + 'chance' + ' ' +
'anti_saliency' + '\n')
KL_itti.write('index' + ' ' + 'saliency' + ' ' + 'chance' + ' ' +
