def group(request, id):
c = {}
group_info = get_group(request.session['token'], id)
c['group_info'] = group_info
c['member_map'] = {member[u'user_id']: member[u'nickname'] for member in group_info[u'members']}
request.session['member_map'] = c['member_map']
if cache.get('group-%s' % id) and settings.CACHE_ENABLED:
c['group'] = cache.get('group-%s' % id)
return render(request, 'group.html', c)
if request.GET.get('ajaxLoad', '0') == '0' and settings.CACHE_ENABLED:
return render(request, 'group-loader.html', c)
get_attachment = lambda x: x[0].get('url', None) if len(x) else None
try:
group = Group.objects.get(id=id)
msgs = list(Message.objects.filter(group=group).order_by('created'))
if msgs:
after_id = msgs[-1].id
if not str(after_id).isdigit():
after_id = 0
elif str(after_id) != str(group_info[u'messages'][u'last_message_id']):
msgs += [Message(
id=msg[u'id'],
created=datetime.fromtimestamp(float(msg[u'created_at'])),
author=msg[u'user_id'] if msg[u'user_id'] != 'system' else 0,
text=msg[u'text'],
img=get_attachment(msg[u'attachments']),
likes=msg[u'favorited_by'],
n_likes=len(msg[u'favorited_by'])
) for msg in msg_concurrent(request.session['token'], id, after_id=after_id, n_workers=(int(group_info[u'messages'][u'count'])/10 + 1))]
group.analysis = analysis(request, msgs, group_info)
except Group.DoesNotExist:
msgs = [Message(
id=msg[u'id'],
created=datetime.fromtimestamp(float(msg[u'created_at'])),
author=msg[u'user_id'] if msg[u'user_id'] != 'system' else 0,
text=msg[u'text'],
img=get_attachment(msg[u'attachments']),
likes=msg[u'favorited_by'],
n_likes=len(msg[u'favorited_by'])
) for msg in msg_concurrent(request.session['token'], id, n_workers=(int(group_info[u'messages'][u'count'])/10 + 1))]
group = Group(id=id, analysis=analysis(request, msgs, group_info))
def save_msg(m):
m.group = group
m.save()
map(lambda m: save_msg(m), msgs)
if settings.CACHE_ENABLED:
cache.set('group-%s' % id, group, 180)
group.save()
c['group'] = group
return render(request, 'group.html', c)
def analysis(self):
# 計算對中數量
counts = [
analysis(
self.history.iloc[i, 1:13].values.tolist(),
self.deck
) for i in range(self.tableRows)
]
self.history["對中數量"] = counts
def data_stats(self, sample_size):
"""compute sigma and mu of each frequency bin in noise dir, from DeepXi"""
if os.path.exists(self.noise_dir + 'stats.npz'):
with np.load(self.noise_dir + 'stats.npz') as stats:
self.mu = stats['mu_hat']
self.sigma = stats['sigma_hat']
else:
print('Start saving stats')
samples = []
for idx in range(sample_size):
snr = random.choice(self.snr_level)
speech_file = random.choice(self.speech_wav_files)
speech_src, _ = librosa.load(speech_file, sr=self.sr)
noise_file = random.choice(self.noise_wav_files)
noise_src, _ = librosa.load(noise_file, sr=self.sr)
start_idx = random.randint(0, len(noise_src) - len(speech_src))
noise_src = noise_src[start_idx:start_idx + len(speech_src)]
_, alpha = utils.add_noise(
speech_src, noise_src,
snr) # get scale factor based on snr
noise_src = noise_src * alpha
# do stft for both speech and noise
_, sample_speech_mag, _ = utils.analysis(
speech_src, self.frame_len, self.frame_shift, self.n_fft)
_, sample_noise_mag, _ = utils.analysis(
noise_src, self.frame_len, self.frame_shift, self.n_fft)
# compute prior snr between speech and noise spectrums
snr_db = utils.prior_snr(
sample_speech_mag,
sample_noise_mag) # instantaneous a prior SNR (dB).
samples.append(np.squeeze(snr_db))
samples = np.hstack(samples)
if len(samples.shape) != 2:
raise ValueError('Incorrect shape for sample.')
stats = {
'mu_hat': np.mean(samples, axis=1),
'sigma_hat': np.std(samples, axis=1)
}
self.mu, self.sigma = stats['mu_hat'], stats['sigma_hat']
np.savez(self.noise_dir + 'stats.npz',
mu_hat=stats['mu_hat'],
sigma_hat=stats['sigma_hat'])
print('Sample statistics saved.')
def showResult(self):
self.deck = self.deckInput.getValue()
result = self.getNewest()
radNo = result.radno
date = result.date
count = analysis(result(), self.deck)
self.radNoText.SetLabel(str(radNo))
self.dateText.SetLabel(str(date))
for i, t in enumerate(result()):
self.winningNumbers[i].SetLabel(t)
if t in self.deck:
self.winningNumbers[i].SetForegroundColour("#00c853")
else:
self.winningNumbers[i].SetForegroundColour("#000000")
self.oddEvenText.SetLabel(oddAndEven(result()))
self.smallLargeText.SetLabel(smallAndLarge(result()))
self.countText.SetLabel("中了%s個號碼！！" % count)
for index, (img,mask) in enumerate(val_loader):
Model.eval()
DSC = 0
PPV = 0
Sen = 0
batch = 0
img = img.to(device)
mask = mask.to(device)
with torch.no_grad():
output = Model(img)
output = torch.ge(output, 0.5).type(dtype=torch.float32) #二值化
output = utils.post_process(output) #后处理
DSC ,PPV, Sen, batch = utils.analysis(output,mask)
DSC_sum += DSC*batch
PPV_sum += PPV*batch
Sen_sum += Sen*batch
batch_sum += batch
if index%save_step==0:
img_list = [
img[0,:,:,:],
output[0,:,:,:],
mask[0,:,:,:]
]
img_visualize = vutils.make_grid(img_list)
visualize_img_path = visualize_path+str(epoch)+'_'+str(index+1)+'.tif'
vutils.save_image(img_visualize, visualize_img_path)
sys.stdout.write("\r[Val] [Epoch {}/{}] [Batch {}/{}] [DSC:{:.5f}] [PPV:{:.5f}] [Sen:{:.5f}]".format(epoch+1, stop_epoch, index+1, len(val_loader), DSC, PPV, Sen))
sys.stdout.flush()
'レース分析', split_detail[0] + split_detail[1].zfill(2) + split_detail[2].zfill(2)), race + race_name)
os.makedirs(directoryName, exist_ok=True)
print(race)
race_level = getLevel(str(race_data.find(class_='RaceData02')))
with open(os.path.join(directoryName, textFileName), 'w') as f:
f.write(link + '\n')
f.write('\n'.join(link_list))
race_data01_class = str(race_data.find_all(class_='RaceData01')).split('m')[0]
distance = int(race_data01_class[-4:])
course = race_data01_class[-5]
horse_tag_list = race_data.find_all(class_='HorseName')
# horse_sex_list = race_data.find_all(class_='Barei Txt_C')
weight_list = race_data.select('td[class="Txt_C"]')
link = link.replace('shutuba', 'data_top')
average_pace_info = getAveragePaceInfo(directoryName, link_list, distance)
track_list = ['札幌', '函館', '福島', '新潟', '東京', '中山', '中京', '京都', '阪神', '小倉']
with open(os.path.join(directoryName, '平均指数表.csv'), 'w', newline='', encoding='cp932') as f:
writer = csv.writer(f)
writer.writerow(['馬名', '分析レース数', '平均トータル指数', '平均先行指数', '平均スタミナ指数', '平均上がり指数', '高速平均トータル指数', '中速平均トータル指数', '低速平均トータル指数',
'枠複勝率良', '種牡馬複勝率良', '枠複勝率稍', '種牡馬複勝率稍', '枠有利度重', '種牡馬有利度重', '枠有利度不', '種牡馬有利度不'])
for i, horse_tag in enumerate(horse_tag_list):
link = getLink(horse_tag)
horse_name = getTitle(horse_tag)
if i != 0:
# horse_sex = getRowElement(horse_sex_list[i - 1], 0)[1]
horse_weight = getRowElement(weight_list[i - 1], 0)[1]
num = getNum(i, len(horse_tag_list) - 1)
if link[0] and horse_name[0]:
analysis(directoryName, average_pace_info,
course, horse_name[1], link[1], track_list[int(trackId) - 1], num, float(horse_weight), race_level)
data = AbileneMat['P']
n = data.shape[0]
e_high = 0.98
e_low = 0.96
alpha = 0.96
sci = -1
holdOFF = 50
# My version
res_me = FRAHST_V3(data, alpha=0.96, e_low=0.96, e_high=0.98, sci = -1, \
holdOffTime=holdOFF, r = 1, evalMetrics = 'F')
metric_me, sets_me, anom_det_tab_me = analysis(res_me, AbileneMat['P_g_truth_tab_alt'], n)
new_metrics_me = scoreEvaluation(n, res_me['anomalies'],
AbileneMat['P_g_truth_tab_alt'], ignoreUpTo = 400)
# Pedros Version
res_ped = frahst_pedro(data, r = 1, alpha=0.96, energy_low=0.96, energy_high=0.98, \
holdOffTime=holdOFF, evaluateTruth='FALSE')
metric_ped, sets_ped, anom_det_tab_ped = analysis(res_ped, AbileneMat['P_g_truth_tab_alt'], n)
new_metrics_ped = scoreEvaluation(n, res_ped['anomalies'],
AbileneMat['P_g_truth_tab_alt'], ignoreUpTo = 400)