def forward(self, input, spec_aug=False, mixup_lambda=None):
#print(input.type()) # Input : (16, 144000)
x = self.spectrogram_extractor(input.float()) # Output : (batch_size, 1, time_steps, n_fft + 1) : (16, 1, 696, 513)
x = self.logmel_extractor(x) # Output : (batch_size, 1, time_steps, mel_bins) : (16, 1 , 696, 128)
frames_num = x.shape[2]
if self.training:
x = self.spec_augmenter(x)
# Mixup on spectrogram
if mixup_lambda is not None:
x = do_mixup(x, mixup_lambda)
x = x.transpose(1, 3)
x = self.batch_norm(x)
x = x.transpose(1, 3)
# (16, 1, 2087, 128)
x = self.encoder.forward_features(x) # output : (batch_size, n_features, 66, 4)
# Aggregate in time axis
x = torch.mean(x, dim=3) # (16, 2048, 22) : (batch_size, n_features, _)
x1 = F.max_pool1d(x, kernel_size=3, stride=1, padding=1)
x2 = F.avg_pool1d(x, kernel_size=3, stride=1, padding=1)
x = x1 + x2 # (16, 2048, 22)
x = F.dropout(x, p=0.5, training=self.training)
x = x.transpose(1, 2) # (batch_size, 22, n_features)
#x = self.encoder.classifier(x) # (16, 22, 2048) : (batch_size, time, n_features)
x = F.relu_(self.encoder.fc(x)) # (16, 22, 2048)
x = x.transpose(1, 2) # (16, 2048, 22)
x = F.dropout(x, p=0.5, training=self.training)
(clipwise_output, norm_att, segmentwise_output) = self.att_head(x)
segmentwise_output = segmentwise_output.transpose(1, 2)
#print("clipwise_output.size : {}".format(clipwise_output.size())) #(16, 24) : (batch_sizes, n_features)
#print("norm_att.size : {}".format(norm_att.size())) #(16, 24, 22) : (batch_sizes, n_features, time)
#print("segmentwise_output.size : {}".format(segmentwise_output.size())) #(16, 24, 22) : (batch_sizes, n_features, time)
#Upscale back to original size
framewise_output = interpolate(segmentwise_output,
self.interpolate_ratio) # (16,696, 24) : (batch_sizes x time x num_classes)
framewise_output = pad_framewise_output(framewise_output, frames_num) # (16,696, 24) : (batch_sizes x time x num_classes)
output_dict = {
'framewise_output': framewise_output,
'clipwise_output': clipwise_output
}
return output_dict
def preprocess_data():
print('Starting the ETL process')
start_time = time.time()
df_train,df_sub = ETL_emb()
test_lengths = count_test_period(df_sub)
elapsed_time = time.time() - start_time
print('Time taken to complete the ETL process {}'.format(elapsed_time))
forecast_ids = list(test_lengths.keys())
site_ids = list(df_sub.SiteId.unique())
# suppress pandas warnings
# do not do this in production!
pd.set_option('mode.chained_assignment', None)
submission_df = pd.read_csv('../data/forecast/submission_format.csv')
submission_df = submission_df.set_index('obs_id')
data_df = preprocess_seq2seq(df_train,df_sub)
# reset index to make sure we do not have double indexes and for easier indexing
data_df = data_df.reset_index()
del data_df['index']
# leave only the first holiday
data_df = data_df[(data_df['obs_id'].shift(+1) != data_df['obs_id'])]
data_df = data_df.reset_index()
del data_df['index']
# fill days wo holidays with -1
data_df['Holiday'] = data_df['Holiday'].fillna(value=-1)
data_df,train_forecast_ids,normal_forecast_ids,linear_interpolation,last_window,submit_zeroes,submit_averages = interpolate(data_df)
return data_df,train_forecast_ids,normal_forecast_ids,linear_interpolation,last_window,submit_zeroes,submit_averages
deformation = sdn.predict(brain_test)
flip = np.load(LRpath + test_label[:2] + 'LR.npy')
X = flip[..., 0].flatten()
Y = flip[..., 1].flatten()
Z = flip[..., 2].flatten()
# Xnii = sitk.ReadImage(dispath+'S'+test_label[:2]+'to' +'S'+ test_label[2:4]+'.fsl152.skullstripped_DispField_Trg2Tpl_X.nii')
# Ynii = sitk.ReadImage(dispath+'S'+test_label[:2]+'to' +'S'+ test_label[2:4]+'.fsl152.skullstripped_DispField_Trg2Tpl_Y.nii')
# Znii = sitk.ReadImage(dispath+'S'+test_label[:2]+'to' +'S'+ test_label[2:4]+'.fsl152.skullstripped_DispField_Trg2Tpl_Z.nii')
# dispX = sitk.GetArrayViewFromImage(Xnii)
# dispY = sitk.GetArrayViewFromImage(Ynii)
# dispZ = sitk.GetArrayViewFromImage(Znii)
# deformation = np.stack([dispX, dispY, dispZ], 3)
# deformation = np.expand_dims(deformation, 0)
disp_LR = interpolate(deformation, Y, Z, X, (91, 109, 91))
disp_LR = np.reshape(disp_LR, (91, 109, 91, 3))
disp_pf = np.einsum('abcde,abce->abcd', Jac(flip),
disp_LR[:-1, :-1, :-1, :])
disp_N = np.mean(deformation**2)
disp_LR_N = 0.25 * np.mean((deformation[0, :-1, :-1, :-1, :] + disp_pf)**2)
test_log.write('{}\t'.format(test_label))
test_log.write('{:6f}\t'.format(disp_N))
test_log.write('{:6f}\n'.format(disp_LR_N))
print('Test sample {}\'s evaluation completed.'.format(test_ind + 1))
#print([disp_N, disp_LR_N])
test_log.close()