def crop_oasis_images():
# Get the list of files from the source data
files = list_files(SOURCE, EXTENSION)
i = 0
# Iterate over each file path
for f in files:
# Load and crop the image at path f
try:
print i, f
cropped_image = load_and_crop_image(f)
# Assign a path for saving. We need to move it from source to dest,
# so do some path replacement
save_path = f.replace(SOURCE,
DEST).replace(EXTENSION, SAVE_EXTENSION)
i += 1
# Save the new cropped image
plt.imsave(save_path,
cropped_image,
cmap=plt.cm.grey,
format=SAVE_FORMAT)
except Exception:
print 'Problem cropping', f
def process(source_path, target_path, move):
files = common.list_files(source_path)
logger.info(f"{len(files)} pictures found")
count = 1
for file in files:
picture = Picture(file)
logger.info(f"{count}/{len(files)} - {picture.file_name}")
process_picture(picture, target_path, move)
count += 1
def demo_plot_cropped_image():
# Get the list of files from the source data
files = list_files(SOURCE, EXTENSION, result_cap=1)
# Crop the first image
cropped_image = load_and_crop_image(files[0])
# Plot the image using matplotlib
plt.imshow(cropped_image)
plt.show()
def process(source_path, comparison_images_directory):
files = common.list_files(source_path)
logger.info(f"{len(files)} pictures found")
known_faces = faces.load_known_faces(comparison_images_directory)
count = 1
for file in files:
picture = Picture(file)
logger.info(f"{count}/{len(files)} - {picture.file_name}")
faces.recognize_faces(picture, known_faces)
count += 1
def load_audio(path, seg, nw, n_mfcc):
# seg = 4 # 2s分段
# nw = 1024 # 帧长约23ms*4
# n_mfcc = 32 # mfcc 维数
folder_list = common.list_files(path)
data_dict = {}
for folder in folder_list:
species = folder.split('\\')[-1]
wav_list = common.find_ext_files(folder, '.wav')
instance = dataObejct()
for wav in wav_list:
wavname = wav.split('\\')[-1]
wav_data, fs = common.read_wav_file(wav)
# 多通道仅取一通道
if wav_data.ndim > 1:
wav_data = wav_data[:, 0]
wav_data = wav_data.T
ref_value = 2**12 - 1
wav_data = wav_data / ref_value # wave幅值归一化
filter_data = common.butter_worth_filter(wav_data,
cutoff=1000,
fs=fs,
btype='high',
N=8)
seg_mfcc, frame_num = enframe_and_feature_extract(
filter_data, seg, nw, fs, n_mfcc)
# # 输出为[n_mfcc, n_sample]
# mfcc_data = librosa.feature.mfcc(y=filter_data, sr=fs, n_mfcc=n_mfcc, n_fft=nw, hop_length=inc)
instance.append(audio_name=wavname,
origin=filter_data,
frame=seg_mfcc,
frame_num=frame_num)
data_dict[species] = instance
return data_dict
from common import copy_files
from common import list_files_in_subdirectories_by_extension as list_files
# This Script will find all .txt files in folders of the data and store them in the external_raw folder
# Insure that the Source only points to the folder that has Disk files with the PAT records in them
SOURCE = '/Users/gunnarenserro/Desktop/Oasis_data/'
EXSTENSION = '.txt'
DEST = '../../data/external_raw/Pat_data'
files = list_files(SOURCE, EXSTENSION)
if len(files) % 2 == 1:
raise ValueError('Does not have correct amount of files!')
copy_files(files[0::2], DEST)
def load_and_train(path, seg, nw, n_mfcc, save_path):
"""
:param path: 音频存储路径
:param nw: 帧长
:param n_mfcc: mfcc特征维数
:return:
hmm_models, hmm模型对象
test_list，list[instance('frame':分帧数据 ,'origin':原始音频数据（滤波）, 'frame_num':帧数), ]
"""
hmm_models = hmms(n_iter=1000)
list_folders = common.list_files(path)
test_list = []
for i in range(len(list_folders)):
name = list_folders[i].split('\\')[-1]
config_name = os.path.join(path, name + '.json')
with open(config_name, encoding='UTF-8') as file:
config = json.load(file)
n_components = config['n_components']
n_mixs = config['n_mixs']
audio_num_for_train = config['audio num for train']
# audio_num_for_train=10
list_wavs = common.find_ext_files(list_folders[i], ext='.wav')
print('%d = %s num：%d' % (i, list_folders[i], len(list_wavs)))
list_wavs = shuffle_list(list_wavs)
instance = dataObejct()
instance.set_name(name)
debug = 0
for wavname in list_wavs:
if debug >= audio_num_for_train:
break
debug += 1
wav_data, fs = common.read_wav_file(wavname)
# 多通道仅取一通道
if wav_data.ndim > 1:
wav_data = wav_data[:, 0]
wav_data = wav_data.T
ref_value = 2**12 - 1
wav_data = wav_data / ref_value # wave幅值归一化
filter_data = common.butter_worth_filter(wav_data,
cutoff=1000,
fs=fs,
btype='high',
N=8)
seg_mfcc, frame_num = enframe_and_feature_extract(
filter_data, seg, nw, fs, n_mfcc)
# # 输出为[n_mfcc, n_sample]
# mfcc_data = librosa.feature.mfcc(y=filter_data, sr=fs, n_mfcc=n_mfcc, n_fft=nw, hop_length=inc)
instance.append(audio_name=wavname,
origin=filter_data,
frame=seg_mfcc,
frame_num=frame_num)
split_ = int(instance.get_num() / 2)
shuffled_instance = instance.shuffle()
train_samples = shuffled_instance[split_:]
# 确定训练集最大帧数，留作样本平衡使用。
train_samples.recompute_total()
train_samples.set_name(name)
frames = np.empty((0, n_mfcc))
frames_num = []
npy_name = name + '_' + str(n_mixs) + '_' + str(n_components) + '.npy'
save_name = os.path.join(save_path, npy_name)
if not os.path.exists(save_name):
for j in range(len(train_samples.origin)):
frame_data = train_samples.frame[j]
frame_data = frame_data.reshape((-1, n_mfcc))
frames = np.vstack((frames, frame_data))
frames_num += train_samples.frame_num[j]
if sum(frames_num) != frames.shape[0]:
print('sum(frames_num) = ', sum(frames_num))
print('total frames = ', frames.shape[0])
raise ValueError('sum(frames_num) != frames.shape[0]')
hmm_models.train_one(frames, frames_num, n_components, n_mixs,
name)
else:
print('\t模型%s已存在，加载模型' % npy_name)
hmm_models.load_one_model(save_name)
test_samples = shuffled_instance[:split_]
test_samples.set_name(name)
test_list.append(test_samples)
hmm_models.save_model(save_path)
return hmm_models, test_list
def load_optical_data(path, seg, nw, n_mfcc):
"""
:param path: 音频存储路径
:param nw: 帧长
:param n_mfcc: mfcc特征维数
:return:
train_list，list[instance('frame':分帧数据 ,'origin':原始音频数据（滤波）, 'frame_num':帧数), ]
test_list，list[instance('frame':分帧数据 ,'origin':原始音频数据（滤波）, 'frame_num':帧数), ]
"""
list_folders = common.list_files(path)
train_list = []
test_list = []
for i in range(len(list_folders)):
name = list_folders[i].split('\\')[-1]
list_wavs = common.find_ext_files(list_folders[i], ext='.wav')
print('%d = %s num：%d' % (i, list_folders[i], len(list_wavs)))
list_wavs = shuffle_list(list_wavs)
instance = dataObejct()
instance.set_name(name)
debug = 0
for wavname in list_wavs:
if debug >= 200:
break
debug += 1
wav_data, fs = common.read_wav_file(wavname)
# 多通道仅取一通道
if wav_data.ndim > 1:
wav_data = wav_data[:, 0]
wav_data = wav_data.T
ref_value = 2**12 - 1
wav_data = wav_data / ref_value # wave幅值归一化
filter_data = common.butter_worth_filter(wav_data,
cutoff=1000,
fs=fs,
btype='high',
N=8)
seg_mfcc, frame_num = enframe_and_feature_extract(
filter_data, seg, nw, fs, n_mfcc)
# # 输出为[n_mfcc, n_sample]
# mfcc_data = librosa.feature.mfcc(y=filter_data, sr=fs, n_mfcc=n_mfcc, n_fft=nw, hop_length=inc)
instance.append(audio_name=wavname,
origin=filter_data,
frame=seg_mfcc,
frame_num=frame_num)
split_ = int(instance.get_num() / 2)
shuffled_instance = instance.shuffle()
train_samples = shuffled_instance[split_:]
# 确定训练集最大帧数，留作样本平衡使用。
train_samples.recompute_total()
train_samples.set_name(name)
test_samples = shuffled_instance[:split_]
test_samples.set_name(name)
train_list.append(train_samples)
test_list.append(test_samples)
return train_list, test_list
SOURCES = [
'../../data/external_raw/gifs_cropped_cor',
'../../data/external_raw/gifs_cropped_sag',
'../../data/external_raw/gifs_cropped_tra'
]
DESTS = [
'../../data/external_raw/gifs_cropped_cor_scaled',
'../../data/external_raw/gifs_cropped_sag_scaled',
'../../data/external_raw/gifs_cropped_tra_scaled'
]
EXTENSION = ''
SAVE_FORMAT = 'PNG'
for SOURCE, DEST in zip(SOURCES, DESTS):
files = list_files(SOURCE, EXTENSION, result_cap=1000)
heights = []
widths = []
for f in files:
img = io.imread(f)
heights.append(len(img))
widths.append(len(img[0]))
widths = sorted(widths)
heights = sorted(heights)
print 'Width: mean: ', mean(widths), 'std', std(widths)
print 'Height: mean: ', mean(heights), 'std', std(heights)
def process(path_to_clean, reference_path, delete_duplicates):
files_to_clean = common.list_files(path_to_clean)
nb_files_to_clean = len(files_to_clean)
logger.info(f"{nb_files_to_clean} files to check for duplicates")
if delete_duplicates is True:
logger.info("/!\ Duplicates found will be deleted")
if reference_path is not None:
reference_files = common.list_files(reference_path)
else:
reference_files = []
nb_reference_files = len(reference_files)
logger.info(f"{nb_reference_files} reference files to compare")
total_nb_files = nb_files_to_clean + nb_reference_files
duplicates = {}
p_reference_files = {}
with Bar('Processing', max=total_nb_files) as bar:
if reference_path is not None:
for i in range(nb_reference_files):
f = Picture(reference_files[i])
p_reference_files[f.md5sum] = f
bar.next()
for i in range(nb_files_to_clean):
p_file_to_clean = Picture(files_to_clean[i])
is_duplicate(p_reference_files, duplicates, p_file_to_clean)
bar.next()
nb_duplicates = len(duplicates)
if nb_duplicates == 0:
logger.info("No duplicates found")
elif nb_duplicates == 1:
logger.info(f"One picture with duplicates found")
else:
logger.info(f"{nb_duplicates} pictures with duplicates found")
if nb_duplicates > 0:
with open('report.html', 'a') as the_file:
the_file.write('''
<html>
<head>
<style>
table, th, td {
border: 1px solid black;
border-collapse: collapse;
}
</style>
</head>
<body>
<table border="1"><tr><th>Original</th><th>Duplicates</th></tr>\n'''
)
for d in duplicates.values():
#logger.info(f"Original file: {d.original.full_path}")
the_file.write('<tr>\n')
the_file.write(
f'<!--Original--><td><img title="{d.original.full_path}" alt="{d.original.full_path}" src="file://{d.original.full_path}" width="25%" height="25%"></td>'
)
the_file.write('\n<!--Duplicates--><td>')
for v in d.duplicates:
#logger.info(f" - {v.full_path}")
the_file.write(
f'<img alt="{v.full_path}" src="file://{v.full_path}" width="25%" height="25%">\n'
)
if delete_duplicates is True:
logger.info(
f"Deleting {v.full_path} (original: {d.original.full_path})"
)
os.remove(v.full_path)
the_file.write('</td>\n</tr>\n')
the_file.write('</table></body></html>')
parser.add_argument("--input", help="Input file or folder.")
parser.add_argument("--output", help="Output folder.")
parser.add_argument("--fps", type=int, help="")
parser.add_argument("--to-video", help="")
parser.add_argument("--tool", default="vlc", help="")
args = parser.parse_args()
shutil.rmtree(args.output, ignore_errors=True)
Path(args.output).mkdir(parents=True, exist_ok=True)
sample_fps = args.fps
if args.to_video:
shutil.rmtree(args.to_video, ignore_errors=True)
Path(args.to_video).mkdir(parents=True, exist_ok=True)
for full_path, sub_path, file_name, file_ext in tqdm(list_files(
args.input)):
if file_ext in set(['.avi', '.mp4']):
print(f'To images: {full_path}')
video_duration, video_fps = get_video_info(full_path)
step = math.ceil(video_fps / sample_fps)
if args.tool == "ffmpeg":
for s in tqdm(range(int(video_duration))):
for t in range(sample_fps):
st = s + t / sample_fps
command = f'ffmpeg -ss {st} -i {full_path} -qmin 1 -q:v 1 -vframes 1 {args.output}/{file_name}_{st:08.2f}.jpg'
result = run_shell_command(command)
elif args.tool == "vlc":
command = f"vlc {full_path} --intf=dummy --rate=5 --video-filter=scene --vout=dummy --scene-format=jpg \
--scene-ratio={step} --scene-path={args.output} --scene-prefix={file_name}_ vlc://quit"
def demo_bounds():
files = list_files(SOURCE, EXTENSION, result_cap=100)
i = 0
image = io.imread(files[i])
plt.figure(i)
show_plot_with_bounds_for_image(image)