-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathpipeline_oracle_buckeye.sh
134 lines (108 loc) · 8.45 KB
/
pipeline_oracle_buckeye.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
#!/bin/bash
#SBATCH -t 2:00:00
#SBATCH -n 1
#SBATCH --mem 80G
#SBATCH -N 1
id="be1375fy_oracle_brokenstops30mix"
corpora_audio="/om2/user/szhi/corpora/buckeye_segments"
corpora_audio_alignment="/om2/user/szhi/corpora/buckeye_segments"
corpora_video="/om2/user/szhi/corpora/buckeye_synthetic_video"
data_pretrain="be1375fy/pretrain.json"
data_train="be1375fy/train.json"
data_test="be1375fy/test.json"
video_fn="video_transform_7797"
offset=120
frame_fn="frame_transform_offset_frontpad_"$offset
video_n_components=20
dpgmm_alpha=1.0
echo $id
mkdir $id
cd /om2/user/szhi/perceptual-tuning-pnas
source /om2/user/szhi/miniconda3/etc/profile.d/conda.sh
conda activate synthetic_Dataset
# pretrain script
# singularity exec --nv --bind /om2/user/szhi/ /om2/user/szhi/vagrant/wav2lip-local.simg python video/pretrain_visual_features.py $data_pretrain $id/$id.pkl --video_fn $video_fn --n_components $video_n_components --video_corpus $corpora_video
# extract audiovisual features for train and eval
# python video/extract_oracle_features_buckeye.py $data_train be1375fy_baseline/be1375fy_baseline.pkl $id/${id}_train.mat --video_fn $video_fn --frame_fn $frame_fn --audio_corpus $corpora_audio --video_corpus $corpora_video
# python video/extract_oracle_features_buckeye.py $data_test be1375fy_baseline/be1375fy_baseline.pkl $id/${id}_eval.npy --video_fn $video_fn --frame_fn $frame_fn --audio_corpus $corpora_audio --video_corpus $corpora_video --split eval --mode av
# python video/extract_oracle_features_buckeye.py $data_train be1375fy_baseline/be1375fy_baseline.pkl $id/${id}_a_train.mat --video_fn $video_fn --frame_fn $frame_fn --audio_corpus $corpora_audio --video_corpus $corpora_video --mode a
# python video/extract_oracle_features_buckeye.py $data_test be1375fy_baseline/be1375fy_baseline.pkl $id/${id}_a_eval.npy --video_fn $video_fn --frame_fn $frame_fn --audio_corpus $corpora_audio --video_corpus $corpora_video --split eval --mode a
# python video/extract_oracle_features_buckeye.py $data_train be1375fy_baseline/be1375fy_baseline.pkl $id/${id}_v_train.mat --video_fn $video_fn --frame_fn $frame_fn --audio_corpus $corpora_audio --video_corpus $corpora_video --mode v
# python video/extract_oracle_features_buckeye.py $data_test be1375fy_baseline/be1375fy_baseline.pkl $id/${id}_v_eval.npy --video_fn $video_fn --frame_fn $frame_fn --audio_corpus $corpora_audio --video_corpus $corpora_video --split eval --mode v
python video/extract_oracle_features_buckeye.py $data_test be1375fy_baseline/be1375fy_baseline.pkl $id/${id}_an_eval.npy --video_fn $video_fn --frame_fn $frame_fn --audio_corpus $corpora_audio --video_corpus $corpora_video --split eval --mode an
# python video/extract_oracle_features_buckeye.py $data_test be1375fy_baseline/be1375fy_baseline.pkl $id/${id}_vn_eval.npy --video_fn $video_fn --frame_fn $frame_fn --audio_corpus $corpora_audio --video_corpus $corpora_video --split eval --mode vn
conda deactivate
# # train AV dpgmm
# cp $id/${id}_train.mat /om2/user/szhi/perceptual-tuning-pnas/models/dpgmm/CHILDES/features/${id}_train.mat
# mkdir /om2/user/szhi/perceptual-tuning-pnas/models/dpgmm/CHILDES/models/$id
# module load mit/matlab/2020a
# cd /om2/user/szhi/perceptual-tuning-pnas/tools/dpgmm/dpmm_subclusters_2014-08-06/Gaussian/
# matlab -nojvm -r "feat_mat = '/om2/user/szhi/perceptual-tuning-pnas/models/dpgmm/CHILDES/features/${id}_train.mat'; model_dir = '/om2/user/szhi/perceptual-tuning-pnas/models/dpgmm/CHILDES/models/$id'; my_run_dpgmm_subclusters(feat_mat, model_dir, 15, true, $dpgmm_alpha); exit;"
# cd /om2/user/szhi/perceptual-tuning-pnas
# cp /om2/user/szhi/perceptual-tuning-pnas/models/dpgmm/CHILDES/models/$id/1501-final.mat $id/${id}_1501-final.mat
# # train A dpgmm
# cp $id/${id}_a_train.mat /om2/user/szhi/perceptual-tuning-pnas/models/dpgmm/CHILDES/features/${id}_a_train.mat
# mkdir /om2/user/szhi/perceptual-tuning-pnas/models/dpgmm/CHILDES/models/${id}_a
# module load mit/matlab/2020a
# cd /om2/user/szhi/perceptual-tuning-pnas/tools/dpgmm/dpmm_subclusters_2014-08-06/Gaussian/
# matlab -nojvm -r "feat_mat = '/om2/user/szhi/perceptual-tuning-pnas/models/dpgmm/CHILDES/features/${id}_a_train.mat'; model_dir = '/om2/user/szhi/perceptual-tuning-pnas/models/dpgmm/CHILDES/models/${id}_a'; my_run_dpgmm_subclusters(feat_mat, model_dir, 15, true, $dpgmm_alpha); exit;"
# cd /om2/user/szhi/perceptual-tuning-pnas
# cp /om2/user/szhi/perceptual-tuning-pnas/models/dpgmm/CHILDES/models/${id}_a/1501-final.mat $id/${id}_a_1501-final.mat
# # train V dpgmm
# cp $id/${id}_v_train.mat /om2/user/szhi/perceptual-tuning-pnas/models/dpgmm/CHILDES/features/${id}_v_train.mat
# mkdir /om2/user/szhi/perceptual-tuning-pnas/models/dpgmm/CHILDES/models/${id}_v
# module load mit/matlab/2020a
# cd /om2/user/szhi/perceptual-tuning-pnas/tools/dpgmm/dpmm_subclusters_2014-08-06/Gaussian/
# matlab -nojvm -r "feat_mat = '/om2/user/szhi/perceptual-tuning-pnas/models/dpgmm/CHILDES/features/${id}_v_train.mat'; model_dir = '/om2/user/szhi/perceptual-tuning-pnas/models/dpgmm/CHILDES/models/${id}_v'; my_run_dpgmm_subclusters(feat_mat, model_dir, 15, true, $dpgmm_alpha); exit;"
# cd /om2/user/szhi/perceptual-tuning-pnas
# cp /om2/user/szhi/perceptual-tuning-pnas/models/dpgmm/CHILDES/models/${id}_v/1501-final.mat $id/${id}_v_1501-final.mat
### switch to abx_eval conda env ###
conda activate abx_eval
# # convert eval .npy to .features with mat2h5
# python tools/mat2h5.py $id/${id}_eval.npy $id/${id}_eval.features
# # extract_posteriors
# python tools/dpgmm/extract_posteriors.py $id/${id}_eval.features $id/${id}_1501-final.mat $id/${id}_eval_posteriors.features
# # abxpy_eval
# python eval/abxpy_eval_buckeye.py $data_test $id/${id}_eval_posteriors.features $id --av_offset $offset --alignment_corpus $corpora_audio_alignment
# mkdir ${id}_aa
# # convert eval .npy to .features with mat2h5
# python tools/mat2h5.py $id/${id}_a_eval.npy $id/${id}_a_eval.features
# # extract_posteriors
# python tools/dpgmm/extract_posteriors.py $id/${id}_a_eval.features $id/${id}_a_1501-final.mat $id/${id}_aa_eval_posteriors.features
# # abxpy_eval
# python eval/abxpy_eval_buckeye.py $data_test $id/${id}_aa_eval_posteriors.features ${id}_aa --av_offset $offset --alignment_corpus $corpora_audio_alignment
# mkdir ${id}_vv
# # convert eval .npy to .features with mat2h5
# python tools/mat2h5.py $id/${id}_v_eval.npy $id/${id}_v_eval.features
# # extract_posteriors
# python tools/dpgmm/extract_posteriors.py $id/${id}_v_eval.features $id/${id}_v_1501-final.mat $id/${id}_vv_eval_posteriors.features
# # abxpy_eval
# python eval/abxpy_eval_buckeye.py $data_test $id/${id}_vv_eval_posteriors.features ${id}_vv --av_offset $offset --alignment_corpus $corpora_audio_alignment
mkdir ${id}_ava
# convert eval .npy to .features with mat2h5
python tools/mat2h5.py $id/${id}_an_eval.npy $id/${id}_an_eval.features
# extract_posteriors
python tools/dpgmm/extract_posteriors.py $id/${id}_an_eval.features $id/${id}_1501-final.mat $id/${id}_ava_eval_posteriors.features
# abxpy_eval
python eval/abxpy_eval_buckeye.py $data_test $id/${id}_ava_eval_posteriors.features ${id}_ava --av_offset $offset --alignment_corpus $corpora_audio_alignment
# mkdir ${id}_avv
# # convert eval .npy to .features with mat2h5
# python tools/mat2h5.py $id/${id}_vn_eval.npy $id/${id}_vn_eval.features
# # extract_posteriors
# python tools/dpgmm/extract_posteriors.py $id/${id}_vn_eval.features $id/${id}_1501-final.mat $id/${id}_avv_eval_posteriors.features
# # abxpy_eval
# python eval/abxpy_eval_buckeye.py $data_test $id/${id}_avv_eval_posteriors.features ${id}_avv --av_offset $offset --alignment_corpus $corpora_audio_alignment
conda deactivate
conda activate h5f_plt
# python eval/analyze_scores.py $id/${id}.txt $id/${id}_analysis.txt
# python eval/analyze_scores.py ${id}_aa/${id}_aa.txt ${id}_aa/${id}_aa_analysis.txt
# python eval/analyze_scores.py ${id}_vv/${id}_vv.txt ${id}_vv/${id}_vv_analysis.txt
python eval/analyze_scores.py ${id}_ava/${id}_ava.txt ${id}_ava/${id}_ava_analysis.txt
# python eval/analyze_scores.py ${id}_avv/${id}_avv.txt ${id}_avv/${id}_avv_analysis.txt
# python eval/analyze_clusters.py $id/${id}_eval_posteriors.features $data_test $id
# python eval/analyze_clusters.py $id/${id}_aa_eval_posteriors.features $data_test ${id}_aa
# python eval/analyze_clusters.py $id/${id}_vv_eval_posteriors.features $data_test ${id}_vv
python eval/analyze_clusters.py $id/${id}_ava_eval_posteriors.features $data_test ${id}_ava
# python eval/analyze_clusters.py $id/${id}_avv_eval_posteriors.features $data_test ${id}_avv
## ADD EVAL FOR AV-A+, AV-V+
### end abx_eval conda env ###