def cut_event(event_id):
# get event info
[event_loc, pick_dict] = pha_list[event_id]
if len(event_loc)==5:
ot, lat, lon, dep, mag = event_loc
if len(event_loc)==4:
ot, lat, lon, mag = event_loc
dep = -1
data_dict = get_data_dict(ot, data_dir)
event_name = dtime2str(ot)
event_dir = os.path.join(out_root, event_name)
if not os.path.exists(event_dir): os.makedirs(event_dir)
# cut event
print('cutting {}'.format(event_name))
for net_sta, [tp, ts] in pick_dict.items():
time_range = [tp-event_win[0], tp+event_win[1]]
t0 = event_win[0]
t1 = ts - tp + event_win[0] if ts else None
if net_sta not in data_dict: continue
data_paths = data_dict[net_sta]
chn_codes = [data_path.split('.')[3][0:3] for data_path in data_paths]
out_paths = [os.path.join(event_dir,'%s.%s.sac'%(net_sta,chn)) for chn in chn_codes]
# cut event
for i in range(3):
st = read(data_paths[i], starttime=time_range[0], endtime=time_range[1])
st.write(out_paths[i])
sac.ch_event(out_paths[i], lon, lat, dep, mag, [t0,t1])
def __getitem__(self, index):
train_paths_i, valid_paths_i = [], []
# get one sta-date
sta_date, samples = self.sta_date_items[index]
net_sta, date = sta_date.split('_')
net, sta = net_sta.split('.')
date = UTCDateTime(date)
dtype = [('tp', 'O'), ('ts', 'O')]
picks = np.array(
[tuple(sample[-3:-1]) for sample in samples if sample[-1]],
dtype=dtype)
# read & prep one day's data
print('reading %s %s' % (net_sta, date.date))
data_dict = get_data_dict(date, data_dir)
if net_sta not in data_dict: return train_paths_i, valid_paths_i
st_paths = data_dict[net_sta]
try:
stream = read(st_paths[0])
stream += read(st_paths[1])
stream += read(st_paths[2])
except:
return train_paths_i, valid_paths_i
if to_filter: stream = preprocess(stream, samp_rate, freq_band)
if len(stream) != 3: return train_paths_i, valid_paths_i
for [samp_class, event_name, tp, ts, _] in samples:
out_dir = os.path.join(out_root, samp_class, 'negative',
event_name)
samp_name = 'neg_%s_%s_%s' % (net, sta, event_name[:-3])
n_aug = num_aug if samp_class == 'train' else 1
for aug_idx in range(n_aug):
# rand time shift
start_time = tp + np.random.rand(1)[0] * rand_dt
end_time = start_time + win_len
# check if tp-ts exists in selected win
is_tp = (picks['tp'] > max(ts, start_time)) * (picks['tp'] <
end_time)
is_ts = (picks['ts'] > max(ts, start_time)) * (picks['ts'] <
end_time)
if sum(is_tp * is_ts) > 0: continue
# slice & prep
st = obspy_slice(stream, start_time, end_time)
if 0 in st.max() or len(st) != 3: continue
st = st.detrend('demean').normalize(
global_max=global_max_norm) # note: no detrend here
# write & record out_paths
if samp_class == 'train': train_paths_i.append([])
if samp_class == 'valid': valid_paths_i.append([])
for tr in st:
out_path = os.path.join(
out_dir,
'%s.%s.%s' % (aug_idx, samp_name, tr.stats.channel))
tr.stats.sac.t1 = ts - start_time
tr.write(out_path, format='sac')
if samp_class == 'train':
train_paths_i[-1].append(out_path)
if samp_class == 'valid':
valid_paths_i[-1].append(out_path)
return train_paths_i, valid_paths_i
def __getitem__(self, index):
train_paths_i, valid_paths_i = [], []
# get one sta-date
sta_date, samples = self.sta_date_items[index]
net_sta, date = sta_date.split('_')
net, sta = net_sta.split('.')
date = UTCDateTime(date)
# read & prep one day's data
print('reading %s %s' % (net_sta, date.date))
data_dict = get_data_dict(date, data_dir)
if net_sta not in data_dict: return train_paths_i, valid_paths_i
st_paths = data_dict[net_sta]
try:
stream = read(st_paths[0])
stream += read(st_paths[1])
stream += read(st_paths[2])
except:
return train_paths_i, valid_paths_i
if to_filter: stream = preprocess(stream, samp_rate, freq_band)
if len(stream) != 3: return train_paths_i, valid_paths_i
for [samp_class, event_name, tp, ts] in samples:
out_dir = os.path.join(out_root, samp_class, 'positive',
event_name)
samp_name = 'pos_%s_%s_%s' % (net, sta, event_name[:-3])
n_aug = num_aug if samp_class == 'train' else 1
for aug_idx in range(n_aug):
# rand time shift & prep
start_time = tp - np.random.rand(1)[0] * rand_dt
end_time = start_time + win_len
st = obspy_slice(stream, start_time, end_time)
if 0 in st.max() or len(st) != 3: continue
st = st.detrend('demean').normalize(
global_max=global_max_norm) # note: no detrend here
# write & record out_paths
if samp_class == 'train': train_paths_i.append([])
if samp_class == 'valid': valid_paths_i.append([])
for tr in st:
if aug_idx > 0 and max_noise > 0:
tr = add_noise(tr, tp, ts)
out_path = os.path.join(
out_dir,
'%s.%s.%s' % (aug_idx, samp_name, tr.stats.channel))
tr.stats.sac.t0, tr.stats.sac.t1 = tp - start_time, ts - start_time
tr.write(out_path, format='sac')
if samp_class == 'train':
train_paths_i[-1].append(out_path)
if samp_class == 'valid':
valid_paths_i[-1].append(out_path)
return train_paths_i, valid_paths_i
def __getitem__(self, index):
data_paths_i = []
# get one sta-date
sta_date, samples = self.sta_date_items[index]
net_sta, date = sta_date.split('_')
net, sta = net_sta.split('.')
date = UTCDateTime(date)
# read & prep one day's data
print('reading %s %s' % (net_sta, date.date))
data_dict = get_data_dict(date, data_dir)
if net_sta not in data_dict: return data_paths_i
st_paths = data_dict[net_sta]
try:
stream = read(st_paths[0])
stream += read(st_paths[1])
stream += read(st_paths[2])
except:
return data_paths_i
stream = preprocess(stream, samp_rate, freq_band)
if len(stream) != 3: return data_paths_i
for [event_dir, tp, ts] in samples:
# time shift & prep
start_time = tp - win_len[0]
end_time = tp + win_len[1]
if data_format == 'sac':
st = sac.obspy_slice(stream, start_time, end_time)
else:
st = stream.slice(start_time, end_time)
if 0 in st.max() or len(st) != 3: continue
st = st.detrend('demean') # note: no detrend here
# write & record out_paths
data_paths_i.append([])
for tr in st:
out_path = os.path.join(event_dir,
'%s.%s' % (net_sta, tr.stats.channel))
tr.write(out_path, format='sac')
tr = read(out_path)[0]
tr.stats.sac.t0, tr.stats.sac.t1 = tp - start_time, ts - start_time
tr.write(out_path, format='sac')
data_paths_i[-1].append(out_path)
return data_paths_i
def cut_event(event_id):
# get event info
[event_loc, pick_dict] = pha_list[event_id]
ot, lat, lon, dep, mag = event_loc
data_dict = get_data_dict(ot, data_dir)
event_name = dtime2str(ot)
event_dir = os.path.join(out_root, event_name)
if not os.path.exists(event_dir): os.makedirs(event_dir)
# cut event
print('cutting {}'.format(event_name))
for net_sta, [tp, ts] in pick_dict.items():
if net_sta not in data_dict: continue
for data_path in data_dict[net_sta]:
b = tp - read(data_path)[0].stats.starttime - event_win[0]
chn_code = data_path.split('.')[-2]
out_path = os.path.join(event_dir,'%s.%s'%(net_sta,chn_code))
# cut event
sac.cut(data_path, b, b+event_win[1], out_path)
# write header
tn = {}
tn['t0'] = event_win[0]
if ts: tn['t1'] = ts - tp + event_win[0]
sac.ch_event(out_path, lat, lon, dep, mag, tn)