kyakuno
commented
10 months ago 1khzのsinを入れた場合のpythonの期待値。
decoder.argmax
Audio shape : (1, 2048)
Frames.shape (13, 1024)
Frames [[ 4.0745363e-10 4.0745363e-10 4.0745363e-10 ... -1.8477590e+00
-1.4142135e+00 -7.6536685e-01]
[ 3.5565467e-10 3.5565467e-10 3.5565467e-10 ... -1.6128563e+00
-1.2344267e+00 -6.6806692e-01]
[ 3.1963293e-10 3.1963293e-10 3.1963293e-10 ... -1.4495014e+00
-1.1094004e+00 -6.0040313e-01]
...
[ 2.9749622e-10 5.5894488e-01 1.0327955e+00 ... 2.9756206e-10
2.9756206e-10 2.9756206e-10]
[ 3.2590958e-10 6.1229348e-01 1.1313708e+00 ... 3.2596290e-10
3.2596290e-10 3.2596290e-10]
[ 3.6429737e-10 6.8456489e-01 1.2649111e+00 ... 3.6443759e-10
3.6443759e-10 3.6443759e-10]]
Probabilities.shape (13, 360)
Probabilities [[0.00020502 0.00031751 0.00042388 ... 0.0071491 0.01173246 0.01672114]
[0.00038446 0.00050728 0.00061423 ... 0.01150815 0.01685008 0.02029349]
[0.00043563 0.00062878 0.00082959 ... 0.02076474 0.03138043 0.03670036]
...
[0.00017953 0.00025119 0.00032758 ... 0.03704717 0.05419874 0.05310791]
[0.00030062 0.00044596 0.00061663 ... 0.04476431 0.06290425 0.06596494]
[0.00035007 0.00050137 0.00067456 ... 0.04698664 0.06278921 0.06866229]]
bins [[298 298 298 298 298 298 298 298 298 298 298 298 298]]
cents [[7957.37940844 7957.37940844 7957.37940844 7957.37940844 7957.37940844
7957.37940844 7957.37940844 7957.37940844 7957.37940844 7957.37940844
7957.37940844 7957.37940844 7957.37940844]]
dither_cents [[7953.61208946 7954.14764803 7949.19239862 7955.28112399 7966.55823823
7963.17041572 7958.84505225 7963.19714456 7961.19396268 7940.6826132
7955.4193838 7950.37785882 7941.15170714]]
freq [[992.27519294 994.3639072 994.48941636 990.90872761 988.75016576
987.85706148 994.1876741 984.65056225 993.77473096 991.06808855
995.5121274 983.2524569 995.69439125]]
pd : [[0.30528295 0.55763453 0.64612538 0.58593071 0.56871527 0.56871527
0.56871527 0.56871527 0.56871527 0.56871527 0.59678572 0.5792219
0.54140884]]
f0 : [[988.45554384 989.68189213 988.54862566 986.27044549 993.53983795
985.11663805 982.49210512 989.97455928 989.54480726 994.11474896
987.21039173 989.97930728 992.20686384]]
[W NNPACK.cpp:64] Could not initialize NNPACK! Reason: Unsupported hardware.
median pd : [[0.30528295 0.55763453 0.5859307 0.5859307 0.5687153 0.5687153
0.5687153 0.5687153 0.5687153 0.5687153 0.5792219 0.5792219
0.54140884]]
mean f0 : [[989.0687 988.8953 988.167 989.45294 988.309 987.04956 985.8611
987.33716 991.21136 990.29 990.4348 989.7988 991.0931 ]]
final f0 : [989.0687 988.8953 988.167 989.45294 988.309 987.04956 985.8611
987.33716 991.21136 990.29 990.4348 989.7988 991.0931 ]decoder.viterbi
Audio shape : (1, 2048)
pd : [[0.30528295 0.55763453 0.64612538 0.58593071 0.56871527 0.56871527
0.56871527 0.56871527 0.56871527 0.56871527 0.59678572 0.5792219
0.54140884]]
f0 : [[997.74597251 997.24236973 996.38860264 987.06377319 982.97850302
986.79590645 990.38032484 987.02009161 995.42629698 991.83966575
993.10933922 992.50616708 994.77689137]]
[W NNPACK.cpp:64] Could not initialize NNPACK! Reason: Unsupported hardware.
median pd : [[0.30528295 0.55763453 0.5859307 0.5859307 0.5687153 0.5687153
0.5687153 0.5687153 0.5687153 0.5687153 0.5792219 0.5792219
0.54140884]]
mean f0 : [[997.49414 997.1257 993.56494 988.8103 985.61273 986.71826 988.0654
990.9422 991.42865 993.45844 992.48505 993.4641 993.64154]]
final f0 : [997.49414 997.1257 993.56494 988.8103 985.61273 986.71826 988.0654
990.9422 991.42865 993.45844 992.48505 993.4641 993.64154]
CREPEを取り込んでf0に対応する。 https://github.com/axinc-ai/ailia-models/issues/1201