m/2015/okada/diary/2016-11-16
をテンプレートにして作成
[
トップ
] [
新規
|
一覧
|
単語検索
|
最終更新
|
ヘルプ
|
ログイン
]
開始行:
#contents
*データセットの使用方法(整理) [#wcddf599]
-データ作成を行う時の岡田のメモとしています。&br;
&br;
https://drive.google.com/drive/u/1/folders/0B8aLPuELdoTRZ...
上記が論文用にまとめた文章です。&br;
*1000Classの実験 [#z9ec919a]
**1000Class_CLS_LOC [#m1ed2eb6]
**1000Class_LOC [#e44054ba]
提供されている学習データ数 : 544546枚&br;
提供されている検証データ数 : 50000枚&br;
&br;
実験で使用する学習に使用するデータをLdata,検証に使用するL...
提供されている学習データを494546枚と50000枚に分割&br;
-Ldataを494546枚、LOC_Vdataを50000枚とする。
1000Class_CLSの学習データとして提供されている1281167枚か...
CLS_Vdataとする。&br;
&br;
Ldata(データの除外処理前) : 494546枚&br;
CLS_Vdata : 50000枚 &br;
LOC_Vdata : 50000枚 &br;
&br;
Ldata :
#pre{{
nb: [ 310 450 2857 991 2104 787 121 44 27 11 ...
rb: [1 1 1 1 1 1 1 1 1 1 1 1 0 1]
nbatch [ 311 451 2858 992 2105 788 122 45 28 12...
sum_nb 7720
sum_rb 13
sum_nbatch 7733
nMdata: 494080
rMdata: 466
ndata: 494546
aratio > 4.0 : 316
w <= 25 , h <= 25 : 12
Total : 323
Using Data num: 494223
}}
#pre{{
0.00 < rate <= 0.10 | 48864 |
0.10 < rate <= 0.20 | 47579 |
0.20 < rate <= 0.30 | 48838 |
0.30 < rate <= 0.40 | 49668 |
0.40 < rate <= 0.50 | 50274 |
0.50 < rate <= 0.60 | 49509 |
0.60 < rate <= 0.70 | 47808 |
0.70 < rate <= 0.80 | 44872 |
0.80 < rate <= 0.90 | 41568 |
0.90 < rate <= 1.00 | 65243 |
sum : 494223
0.00 < rate <= 0.01 | 4336 |
0.01 < rate <= 0.02 | 5279 |
0.02 < rate <= 0.03 | 5096 |
0.03 < rate <= 0.04 | 5042 |
0.04 < rate <= 0.05 | 4953 |
0.05 < rate <= 0.06 | 4939 |
0.06 < rate <= 0.07 | 4868 |
0.07 < rate <= 0.08 | 4888 |
0.08 < rate <= 0.09 | 4700 |
0.09 < rate <= 0.10 | 4763 |
sum : 48864
rate == 0.0 | 21 |
0.000 < rate <= 0.001 | 195 |
0.001 < rate <= 0.002 | 360 |
0.002 < rate <= 0.003 | 424 |
0.003 < rate <= 0.004 | 466 |
0.004 < rate <= 0.005 | 430 |
0.005 < rate <= 0.006 | 469 |
0.006 < rate <= 0.007 | 522 |
0.007 < rate <= 0.008 | 450 |
0.008 < rate <= 0.009 | 505 |
0.009 < rate <= 0.010 | 515 |
sum : 4336
}}
#pre{{
nb: [ 310 450 2857 991 2104 787 121 44 27 11 ...
rb: [1 1 1 1 1 1 1 1 1 1 1 1 0 0]
nbatch [ 311 451 2858 992 2105 788 122 45 28 12...
sum_nb 7716
sum_rb 12
sum_nbatch 7728
nMdata: 493824
rMdata: 378
ndata: 494202
aratio > 4.0 : 0
w <= 25 , h <= 25 : 0
Total : 0
Using Data num: 494202
0.6
max: 0.942857142857
median: 0.388888888889
mean: 0.403341083716
std: 0.106922416642
min: 0.027972027972
rate == 0 : 0
0.0 < rate <= 0.1 | 19 |
0.1 < rate <= 0.2 | 468 |
0.2 < rate <= 0.3 | 90698 |
0.3 < rate <= 0.4 | 178866 |
0.4 < rate <= 0.5 | 126206 |
0.5 < rate <= 0.6 | 74313 |
0.6 < rate <= 0.7 | 21653 |
0.7 < rate <= 0.8 | 1320 |
0.8 < rate <= 0.9 | 657 |
0.9 < rate <= 1.0 | 2 |
sum : 494202
}}
LOC_Vdata :
#pre{{
nb: [ 33 46 288 97 211 81 12 4 2 1 1 0 0 ...
rb: [1 1 1 1 1 1 1 1 1 1 1 1 1 1]
nbatch [ 34 47 289 98 212 82 13 5 3 2 2 1 ...
sum_nb 776
sum_rb 14
sum_nbatch 790
nMdata: 49664
rMdata: 336
ndata: 50000
aratio > 4.0 : 25
w <= 25 , h <= 25 : 0
Total : 25
Using Data num: 49975
}}
w <= 25,h <= 25のデータ、aratio > 4.0のデータ除外後
#pre{{
Using Data num: 49975
0.00 < rate <= 0.10 | 4953 |
0.10 < rate <= 0.20 | 4730 |
0.20 < rate <= 0.30 | 4937 |
0.30 < rate <= 0.40 | 5074 |
0.40 < rate <= 0.50 | 5128 |
0.50 < rate <= 0.60 | 4990 |
0.60 < rate <= 0.70 | 4881 |
0.70 < rate <= 0.80 | 4538 |
0.80 < rate <= 0.90 | 4189 |
0.90 < rate <= 1.00 | 6555 |
sum : 49975
0.00 < rate <= 0.01 | 438 |
0.01 < rate <= 0.02 | 495 |
0.02 < rate <= 0.03 | 532 |
0.03 < rate <= 0.04 | 486 |
0.04 < rate <= 0.05 | 503 |
0.05 < rate <= 0.06 | 502 |
0.06 < rate <= 0.07 | 505 |
0.07 < rate <= 0.08 | 481 |
0.08 < rate <= 0.09 | 522 |
0.09 < rate <= 0.10 | 489 |
sum : 4953
rate == 0.0 | 3 |
0.000 < rate <= 0.001 | 26 |
0.001 < rate <= 0.002 | 32 |
0.002 < rate <= 0.003 | 50 |
0.003 < rate <= 0.004 | 56 |
0.004 < rate <= 0.005 | 36 |
0.005 < rate <= 0.006 | 54 |
0.006 < rate <= 0.007 | 39 |
0.007 < rate <= 0.008 | 47 |
0.008 < rate <= 0.009 | 50 |
0.009 < rate <= 0.010 | 48 |
sum : 438
}}
データの選別完了後(rate = 0.0のデータを除外)
#pre{{
nb: [ 33 46 288 97 211 81 12 4 2 1 1 0 0 ...
rb: [1 1 1 1 1 1 1 1 1 1 1 1 0 0]
nbatch [ 34 47 289 98 212 82 13 5 3 2 2 1 ...
sum_nb 776
sum_rb 12
sum_nbatch 788
nMdata: 49664
rMdata: 308
ndata: 49972
aratio > 4.0 : 0
w <= 25 , h <= 25 : 0
Total : 0
Using Data num: 49972
}}
教師のしきい値 0.6
#pre{{
0.6
max: 0.9
median: 0.385714285714
mean: 0.402682004781
std: 0.106847190335
min: 0.0571428571429
rate == 0 : 0
0.0 < rate <= 0.1 | 5 |
0.1 < rate <= 0.2 | 35 |
0.2 < rate <= 0.3 | 9299 |
0.3 < rate <= 0.4 | 18018 |
0.4 < rate <= 0.5 | 12779 |
0.5 < rate <= 0.6 | 7524 |
0.6 < rate <= 0.7 | 2099 |
0.7 < rate <= 0.8 | 143 |
0.8 < rate <= 0.9 | 70 |
0.9 < rate <= 1.0 | 0 |
sum : 49972
}}
CLS_Vdata
#pre{{
nb: [ 36 53 325 97 184 65 10 3 2 0 1 0 0 ...
rb: [1 1 1 1 1 1 1 1 1 1 1 1 1 1]
nbatch [ 37 54 326 98 185 66 11 4 3 1 2 1 ...
sum_nb 776
sum_rb 14
sum_nbatch 790
nMdata: 49664
rMdata: 336
ndata: 50000
aratio > 4.0 : 14
w <= 25 , h <= 25 : 0
Total : 14
Using Data num: 49986
}}
#pre{{
nb: [ 36 53 325 97 184 65 10 3 2 0 1 0 0 ...
rb: [1 1 1 1 1 1 1 1 1 1 1 1 0 0]
nbatch [ 37 54 326 98 185 66 11 4 3 1 2 1 ...
sum_nb 776
sum_rb 12
sum_nbatch 788
nMdata: 49664
rMdata: 322
ndata: 49986
aratio > 4.0 : 0
w <= 25 , h <= 25 : 0
Total : 0
Using Data num: 49986
0.5
max: 1.0
median: 1.0
mean: 1.0
std: 0.0
min: 1.0
rate == 0 : 0
0.0 < rate <= 0.1 | 0 |
0.1 < rate <= 0.2 | 0 |
0.2 < rate <= 0.3 | 0 |
0.3 < rate <= 0.4 | 0 |
0.4 < rate <= 0.5 | 0 |
0.5 < rate <= 0.6 | 0 |
0.6 < rate <= 0.7 | 0 |
0.7 < rate <= 0.8 | 0 |
0.8 < rate <= 0.9 | 0 |
0.9 < rate <= 1.0 | 49986 |
sum : 49986
}}
*100Classの実験[#k88fc295]
--以前から使用している100クラス(無作為に選定した)のデータ...
今回の実験は予備実験であるため、Testデータは用いません。&...
&br;
CLS+LOCのデータ&br;
学習データ : 1000Class_LOCの学習用データから抽出した
LOCにおける検証データ : 1000Class_LOCの検証用データから抽...
CLSにおける検証データ : 1000Class_CLSの検証用データたから...
&br;
MCost TITANX(Pascal)で100Class_LOC_CLSで実行結果をとった...
200epochにおける実行結果が457m12.697s &br;
1000Class LOCのデータは大体45万枚程度&br;
100Class LOC_CLSの場合を10万枚だと見積もると約4.5倍実行時...
457m * 4.5 = 2056m = 約34h = 1d10hくらいが見積もり。&br;
&br;
1000Class CLSの場合だと、&br;
457m * 10 = 4570m = 約76h = 3d4hくらいが見積もり。&br;
&br;
あくまで、TITANX(Pascal)を使用した場合であるため、他の場...
***100Class_CLS_LOC_Learningのデータの選別前情報 [#q88ae9...
#pre{{
nb: [ 76 117 740 242 470 160 23 9 5 2 1 0 0 ...
rb: [1 1 1 1 1 1 1 1 1 1 1 1 1 1]
nbatch [ 77 118 741 243 471 161 24 10 6 3 2 1 ...
sum_nb 1845
sum_rb 14
sum_nbatch 1859
nMdata: 118080
rMdata: 437
ndata: 118517
aratio > 4.0 : 42
w <= 25 , h <= 25 : 3
Total : 43
Using Data num: 118474
}}
***100Class_CLS_LOC_Learningのデータの選別後情報 [#e06058...
#pre{{
In [1]: %run CLS_LOC_100c_utilMakeBox.py
nb: [ 76 117 740 242 470 160 23 9 5 2 1 0 0 ...
rb: [1 1 1 1 1 1 1 1 1 1 1 1 0 0]
nbatch [ 77 118 741 243 471 161 24 10 6 3 2 1 ...
sum_nb 1845
sum_rb 12
sum_nbatch 1857
nMdata: 118080
rMdata: 394
ndata: 118474
aratio > 4.0 : 0
w <= 25 , h <= 25 : 0
Total : 0
Using Data num: 118474
0.00 < rate <= 0.10 | 4618 |
0.10 < rate <= 0.20 | 5080 |
0.20 < rate <= 0.30 | 5098 |
0.30 < rate <= 0.40 | 5059 |
0.40 < rate <= 0.50 | 5234 |
0.50 < rate <= 0.60 | 5034 |
0.60 < rate <= 0.70 | 4913 |
0.70 < rate <= 0.80 | 4632 |
0.80 < rate <= 0.90 | 4273 |
0.90 < rate <= 1.00 | 74533 |
sum : 118474
0.00 < rate <= 0.01 | 251 |
0.01 < rate <= 0.02 | 453 |
0.02 < rate <= 0.03 | 454 |
0.03 < rate <= 0.04 | 514 |
0.04 < rate <= 0.05 | 476 |
0.05 < rate <= 0.06 | 491 |
0.06 < rate <= 0.07 | 532 |
0.07 < rate <= 0.08 | 488 |
0.08 < rate <= 0.09 | 462 |
0.09 < rate <= 0.10 | 497 |
sum : 4618
rate == 0.0 | 0 |
0.000 < rate <= 0.001 | 3 |
0.001 < rate <= 0.002 | 12 |
0.002 < rate <= 0.003 | 10 |
0.003 < rate <= 0.004 | 25 |
0.004 < rate <= 0.005 | 21 |
0.005 < rate <= 0.006 | 27 |
0.006 < rate <= 0.007 | 37 |
0.007 < rate <= 0.008 | 32 |
0.008 < rate <= 0.009 | 53 |
0.009 < rate <= 0.010 | 31 |
sum : 251
}}
***教師のしきい値 0.6 [#l26b7045]
#pre{{
0.6
max: 1.0
median: 1.0
mean: 0.745375076996
std: 0.303550180092
min: 0.0559440559441
rate == 0 : 0
0.0 < rate <= 0.1 | 1 |
0.1 < rate <= 0.2 | 37 |
0.2 < rate <= 0.3 | 9232 |
0.3 < rate <= 0.4 | 18505 |
0.4 < rate <= 0.5 | 12937 |
0.5 < rate <= 0.6 | 7400 |
0.6 < rate <= 0.7 | 2182 |
0.7 < rate <= 0.8 | 119 |
0.8 < rate <= 0.9 | 49 |
0.9 < rate <= 1.0 | 68012 |
sum : 118474
}}
***100Class_CLS_Validationの選別後情報(選別前情報と一緒) ...
#pre{{
In [1]: %run CLS_LOC_100c_utilMakeBox.py
nb: [ 3 5 33 9 17 5 0 0 0 0 0 0 0 0]
rb: [1 1 1 1 1 1 1 1 1 1 1 1 0 0]
nbatch [ 4 6 34 10 18 6 1 1 1 1 1 1 0 0]
sum_nb 72
sum_rb 12
sum_nbatch 84
nMdata: 4608
rMdata: 363
ndata: 4971
}}
終了行:
#contents
*データセットの使用方法(整理) [#wcddf599]
-データ作成を行う時の岡田のメモとしています。&br;
&br;
https://drive.google.com/drive/u/1/folders/0B8aLPuELdoTRZ...
上記が論文用にまとめた文章です。&br;
*1000Classの実験 [#z9ec919a]
**1000Class_CLS_LOC [#m1ed2eb6]
**1000Class_LOC [#e44054ba]
提供されている学習データ数 : 544546枚&br;
提供されている検証データ数 : 50000枚&br;
&br;
実験で使用する学習に使用するデータをLdata,検証に使用するL...
提供されている学習データを494546枚と50000枚に分割&br;
-Ldataを494546枚、LOC_Vdataを50000枚とする。
1000Class_CLSの学習データとして提供されている1281167枚か...
CLS_Vdataとする。&br;
&br;
Ldata(データの除外処理前) : 494546枚&br;
CLS_Vdata : 50000枚 &br;
LOC_Vdata : 50000枚 &br;
&br;
Ldata :
#pre{{
nb: [ 310 450 2857 991 2104 787 121 44 27 11 ...
rb: [1 1 1 1 1 1 1 1 1 1 1 1 0 1]
nbatch [ 311 451 2858 992 2105 788 122 45 28 12...
sum_nb 7720
sum_rb 13
sum_nbatch 7733
nMdata: 494080
rMdata: 466
ndata: 494546
aratio > 4.0 : 316
w <= 25 , h <= 25 : 12
Total : 323
Using Data num: 494223
}}
#pre{{
0.00 < rate <= 0.10 | 48864 |
0.10 < rate <= 0.20 | 47579 |
0.20 < rate <= 0.30 | 48838 |
0.30 < rate <= 0.40 | 49668 |
0.40 < rate <= 0.50 | 50274 |
0.50 < rate <= 0.60 | 49509 |
0.60 < rate <= 0.70 | 47808 |
0.70 < rate <= 0.80 | 44872 |
0.80 < rate <= 0.90 | 41568 |
0.90 < rate <= 1.00 | 65243 |
sum : 494223
0.00 < rate <= 0.01 | 4336 |
0.01 < rate <= 0.02 | 5279 |
0.02 < rate <= 0.03 | 5096 |
0.03 < rate <= 0.04 | 5042 |
0.04 < rate <= 0.05 | 4953 |
0.05 < rate <= 0.06 | 4939 |
0.06 < rate <= 0.07 | 4868 |
0.07 < rate <= 0.08 | 4888 |
0.08 < rate <= 0.09 | 4700 |
0.09 < rate <= 0.10 | 4763 |
sum : 48864
rate == 0.0 | 21 |
0.000 < rate <= 0.001 | 195 |
0.001 < rate <= 0.002 | 360 |
0.002 < rate <= 0.003 | 424 |
0.003 < rate <= 0.004 | 466 |
0.004 < rate <= 0.005 | 430 |
0.005 < rate <= 0.006 | 469 |
0.006 < rate <= 0.007 | 522 |
0.007 < rate <= 0.008 | 450 |
0.008 < rate <= 0.009 | 505 |
0.009 < rate <= 0.010 | 515 |
sum : 4336
}}
#pre{{
nb: [ 310 450 2857 991 2104 787 121 44 27 11 ...
rb: [1 1 1 1 1 1 1 1 1 1 1 1 0 0]
nbatch [ 311 451 2858 992 2105 788 122 45 28 12...
sum_nb 7716
sum_rb 12
sum_nbatch 7728
nMdata: 493824
rMdata: 378
ndata: 494202
aratio > 4.0 : 0
w <= 25 , h <= 25 : 0
Total : 0
Using Data num: 494202
0.6
max: 0.942857142857
median: 0.388888888889
mean: 0.403341083716
std: 0.106922416642
min: 0.027972027972
rate == 0 : 0
0.0 < rate <= 0.1 | 19 |
0.1 < rate <= 0.2 | 468 |
0.2 < rate <= 0.3 | 90698 |
0.3 < rate <= 0.4 | 178866 |
0.4 < rate <= 0.5 | 126206 |
0.5 < rate <= 0.6 | 74313 |
0.6 < rate <= 0.7 | 21653 |
0.7 < rate <= 0.8 | 1320 |
0.8 < rate <= 0.9 | 657 |
0.9 < rate <= 1.0 | 2 |
sum : 494202
}}
LOC_Vdata :
#pre{{
nb: [ 33 46 288 97 211 81 12 4 2 1 1 0 0 ...
rb: [1 1 1 1 1 1 1 1 1 1 1 1 1 1]
nbatch [ 34 47 289 98 212 82 13 5 3 2 2 1 ...
sum_nb 776
sum_rb 14
sum_nbatch 790
nMdata: 49664
rMdata: 336
ndata: 50000
aratio > 4.0 : 25
w <= 25 , h <= 25 : 0
Total : 25
Using Data num: 49975
}}
w <= 25,h <= 25のデータ、aratio > 4.0のデータ除外後
#pre{{
Using Data num: 49975
0.00 < rate <= 0.10 | 4953 |
0.10 < rate <= 0.20 | 4730 |
0.20 < rate <= 0.30 | 4937 |
0.30 < rate <= 0.40 | 5074 |
0.40 < rate <= 0.50 | 5128 |
0.50 < rate <= 0.60 | 4990 |
0.60 < rate <= 0.70 | 4881 |
0.70 < rate <= 0.80 | 4538 |
0.80 < rate <= 0.90 | 4189 |
0.90 < rate <= 1.00 | 6555 |
sum : 49975
0.00 < rate <= 0.01 | 438 |
0.01 < rate <= 0.02 | 495 |
0.02 < rate <= 0.03 | 532 |
0.03 < rate <= 0.04 | 486 |
0.04 < rate <= 0.05 | 503 |
0.05 < rate <= 0.06 | 502 |
0.06 < rate <= 0.07 | 505 |
0.07 < rate <= 0.08 | 481 |
0.08 < rate <= 0.09 | 522 |
0.09 < rate <= 0.10 | 489 |
sum : 4953
rate == 0.0 | 3 |
0.000 < rate <= 0.001 | 26 |
0.001 < rate <= 0.002 | 32 |
0.002 < rate <= 0.003 | 50 |
0.003 < rate <= 0.004 | 56 |
0.004 < rate <= 0.005 | 36 |
0.005 < rate <= 0.006 | 54 |
0.006 < rate <= 0.007 | 39 |
0.007 < rate <= 0.008 | 47 |
0.008 < rate <= 0.009 | 50 |
0.009 < rate <= 0.010 | 48 |
sum : 438
}}
データの選別完了後(rate = 0.0のデータを除外)
#pre{{
nb: [ 33 46 288 97 211 81 12 4 2 1 1 0 0 ...
rb: [1 1 1 1 1 1 1 1 1 1 1 1 0 0]
nbatch [ 34 47 289 98 212 82 13 5 3 2 2 1 ...
sum_nb 776
sum_rb 12
sum_nbatch 788
nMdata: 49664
rMdata: 308
ndata: 49972
aratio > 4.0 : 0
w <= 25 , h <= 25 : 0
Total : 0
Using Data num: 49972
}}
教師のしきい値 0.6
#pre{{
0.6
max: 0.9
median: 0.385714285714
mean: 0.402682004781
std: 0.106847190335
min: 0.0571428571429
rate == 0 : 0
0.0 < rate <= 0.1 | 5 |
0.1 < rate <= 0.2 | 35 |
0.2 < rate <= 0.3 | 9299 |
0.3 < rate <= 0.4 | 18018 |
0.4 < rate <= 0.5 | 12779 |
0.5 < rate <= 0.6 | 7524 |
0.6 < rate <= 0.7 | 2099 |
0.7 < rate <= 0.8 | 143 |
0.8 < rate <= 0.9 | 70 |
0.9 < rate <= 1.0 | 0 |
sum : 49972
}}
CLS_Vdata
#pre{{
nb: [ 36 53 325 97 184 65 10 3 2 0 1 0 0 ...
rb: [1 1 1 1 1 1 1 1 1 1 1 1 1 1]
nbatch [ 37 54 326 98 185 66 11 4 3 1 2 1 ...
sum_nb 776
sum_rb 14
sum_nbatch 790
nMdata: 49664
rMdata: 336
ndata: 50000
aratio > 4.0 : 14
w <= 25 , h <= 25 : 0
Total : 14
Using Data num: 49986
}}
#pre{{
nb: [ 36 53 325 97 184 65 10 3 2 0 1 0 0 ...
rb: [1 1 1 1 1 1 1 1 1 1 1 1 0 0]
nbatch [ 37 54 326 98 185 66 11 4 3 1 2 1 ...
sum_nb 776
sum_rb 12
sum_nbatch 788
nMdata: 49664
rMdata: 322
ndata: 49986
aratio > 4.0 : 0
w <= 25 , h <= 25 : 0
Total : 0
Using Data num: 49986
0.5
max: 1.0
median: 1.0
mean: 1.0
std: 0.0
min: 1.0
rate == 0 : 0
0.0 < rate <= 0.1 | 0 |
0.1 < rate <= 0.2 | 0 |
0.2 < rate <= 0.3 | 0 |
0.3 < rate <= 0.4 | 0 |
0.4 < rate <= 0.5 | 0 |
0.5 < rate <= 0.6 | 0 |
0.6 < rate <= 0.7 | 0 |
0.7 < rate <= 0.8 | 0 |
0.8 < rate <= 0.9 | 0 |
0.9 < rate <= 1.0 | 49986 |
sum : 49986
}}
*100Classの実験[#k88fc295]
--以前から使用している100クラス(無作為に選定した)のデータ...
今回の実験は予備実験であるため、Testデータは用いません。&...
&br;
CLS+LOCのデータ&br;
学習データ : 1000Class_LOCの学習用データから抽出した
LOCにおける検証データ : 1000Class_LOCの検証用データから抽...
CLSにおける検証データ : 1000Class_CLSの検証用データたから...
&br;
MCost TITANX(Pascal)で100Class_LOC_CLSで実行結果をとった...
200epochにおける実行結果が457m12.697s &br;
1000Class LOCのデータは大体45万枚程度&br;
100Class LOC_CLSの場合を10万枚だと見積もると約4.5倍実行時...
457m * 4.5 = 2056m = 約34h = 1d10hくらいが見積もり。&br;
&br;
1000Class CLSの場合だと、&br;
457m * 10 = 4570m = 約76h = 3d4hくらいが見積もり。&br;
&br;
あくまで、TITANX(Pascal)を使用した場合であるため、他の場...
***100Class_CLS_LOC_Learningのデータの選別前情報 [#q88ae9...
#pre{{
nb: [ 76 117 740 242 470 160 23 9 5 2 1 0 0 ...
rb: [1 1 1 1 1 1 1 1 1 1 1 1 1 1]
nbatch [ 77 118 741 243 471 161 24 10 6 3 2 1 ...
sum_nb 1845
sum_rb 14
sum_nbatch 1859
nMdata: 118080
rMdata: 437
ndata: 118517
aratio > 4.0 : 42
w <= 25 , h <= 25 : 3
Total : 43
Using Data num: 118474
}}
***100Class_CLS_LOC_Learningのデータの選別後情報 [#e06058...
#pre{{
In [1]: %run CLS_LOC_100c_utilMakeBox.py
nb: [ 76 117 740 242 470 160 23 9 5 2 1 0 0 ...
rb: [1 1 1 1 1 1 1 1 1 1 1 1 0 0]
nbatch [ 77 118 741 243 471 161 24 10 6 3 2 1 ...
sum_nb 1845
sum_rb 12
sum_nbatch 1857
nMdata: 118080
rMdata: 394
ndata: 118474
aratio > 4.0 : 0
w <= 25 , h <= 25 : 0
Total : 0
Using Data num: 118474
0.00 < rate <= 0.10 | 4618 |
0.10 < rate <= 0.20 | 5080 |
0.20 < rate <= 0.30 | 5098 |
0.30 < rate <= 0.40 | 5059 |
0.40 < rate <= 0.50 | 5234 |
0.50 < rate <= 0.60 | 5034 |
0.60 < rate <= 0.70 | 4913 |
0.70 < rate <= 0.80 | 4632 |
0.80 < rate <= 0.90 | 4273 |
0.90 < rate <= 1.00 | 74533 |
sum : 118474
0.00 < rate <= 0.01 | 251 |
0.01 < rate <= 0.02 | 453 |
0.02 < rate <= 0.03 | 454 |
0.03 < rate <= 0.04 | 514 |
0.04 < rate <= 0.05 | 476 |
0.05 < rate <= 0.06 | 491 |
0.06 < rate <= 0.07 | 532 |
0.07 < rate <= 0.08 | 488 |
0.08 < rate <= 0.09 | 462 |
0.09 < rate <= 0.10 | 497 |
sum : 4618
rate == 0.0 | 0 |
0.000 < rate <= 0.001 | 3 |
0.001 < rate <= 0.002 | 12 |
0.002 < rate <= 0.003 | 10 |
0.003 < rate <= 0.004 | 25 |
0.004 < rate <= 0.005 | 21 |
0.005 < rate <= 0.006 | 27 |
0.006 < rate <= 0.007 | 37 |
0.007 < rate <= 0.008 | 32 |
0.008 < rate <= 0.009 | 53 |
0.009 < rate <= 0.010 | 31 |
sum : 251
}}
***教師のしきい値 0.6 [#l26b7045]
#pre{{
0.6
max: 1.0
median: 1.0
mean: 0.745375076996
std: 0.303550180092
min: 0.0559440559441
rate == 0 : 0
0.0 < rate <= 0.1 | 1 |
0.1 < rate <= 0.2 | 37 |
0.2 < rate <= 0.3 | 9232 |
0.3 < rate <= 0.4 | 18505 |
0.4 < rate <= 0.5 | 12937 |
0.5 < rate <= 0.6 | 7400 |
0.6 < rate <= 0.7 | 2182 |
0.7 < rate <= 0.8 | 119 |
0.8 < rate <= 0.9 | 49 |
0.9 < rate <= 1.0 | 68012 |
sum : 118474
}}
***100Class_CLS_Validationの選別後情報(選別前情報と一緒) ...
#pre{{
In [1]: %run CLS_LOC_100c_utilMakeBox.py
nb: [ 3 5 33 9 17 5 0 0 0 0 0 0 0 0]
rb: [1 1 1 1 1 1 1 1 1 1 1 1 0 0]
nbatch [ 4 6 34 10 18 6 1 1 1 1 1 1 0 0]
sum_nb 72
sum_rb 12
sum_nbatch 84
nMdata: 4608
rMdata: 363
ndata: 4971
}}
ページ名:
![[PukiWiki]](image/blackuni3-80x80.png "[PukiWiki]")
