params.cfg

[MODEL_TYPE]
model = 1
w_gan = True
unroll_steps = 5
first_kernel_size = 3
kernel_size = 3
dropout = 0.4

[PATH]
datasets = trueness_moveSpeakerOnly
dir_path = /gpfsdswork/projects/rech/urk/uln35en/Projets/non-verbal-behaviours-generation/
data_path = /gpfsdswork/projects/rech/urk/uln35en/Data/
saved_path = generation/saved_models/
output_path = generation/data/output/
evaluation_path = generation/evaluation/
model_path = 10-05-2023_trueness_moveSpeakerOnly_CGAN_50/

[TRAIN]
n_epochs = 2000
batch_size = 32
d_lr = 1e-5
g_lr = 1e-4
log_interval = 50
adversarial_coeff = 0.1
au_coeff = 1
pose_coeff = 1
eye_coeff = 1
fake_target = True

[DATA]
noise_size = 50
pose_size = 11
eye_size = 8
pose_t_size = 3
pose_r_size = 3
au_size = 17
derivative = True
sequence_len = 100

[selected_opensmile_columns]
0 = Speak
1 = alphaRatio_sma3
2 = hammarbergIndex_sma3
3 = mfcc1_sma3
4 = mfcc2_sma3
5 = mfcc3_sma3
6 = F0semitoneFrom27.5Hz_sma3nz
7 = logRelF0-H1-H2_sma3nz

[opensmile_columns]
0 = Speak
1 = Loudness_sma3
2 = alphaRatio_sma3
3 = hammarbergIndex_sma3
4 = slope0-500_sma3
5 = slope500-1500_sma3
6 = spectralFlux_sma3
7 = mfcc1_sma3
8 = mfcc2_sma3
9 = mfcc3_sma3
10 = mfcc4_sma3
11 = F0semitoneFrom27.5Hz_sma3nz
12 = jitterLocal_sma3nz
13 = shimmerLocaldB_sma3nz
14 = HNRdBACF_sma3nz
15 = logRelF0-H1-H2_sma3nz
16 = logRelF0-H1-A3_sma3nz
17 = F1frequency_sma3nz
18 = F1bandwidth_sma3nz
19 = F1amplitudeLogRelF0_sma3nz
20 = F2frequency_sma3nz
21 = F2bandwidth_sma3nz
22 = F2amplitudeLogRelF0_sma3nz
23 = F3frequency_sma3nz
24 = F3bandwidth_sma3nz
25 = F3amplitudeLogRelF0_sma3nz

[openface_columns]
0 = timestamp
1 = gaze_0_x
2 = gaze_0_y
3 = gaze_0_z
4 = gaze_1_x
5 = gaze_1_y
6 = gaze_1_z
7 = gaze_angle_x
8 = gaze_angle_y
9 = pose_Tx
10 = pose_Ty
11 = pose_Tz
12 = pose_Rx
13 = pose_Ry
14 = pose_Rz
15 = AU01_r
16 = AU02_r
17 = AU04_r
18 = AU05_r
19 = AU06_r
20 = AU07_r
21 = AU09_r
22 = AU10_r
23 = AU12_r
24 = AU14_r
25 = AU15_r
26 = AU17_r
27 = AU20_r
28 = AU23_r
29 = AU25_r
30 = AU26_r
31 = AU45_r