【再々開催】Raspberry_PiとUSBカメラで_はじめようディープラーニング20191207

by Yamada3

Slide 1

Slide 1 text

3BTQCFSSZ1Jͱ 64#ΧϝϥͰ͸͡ΊΑ͏ σΟʔϓϥʔχϯά

Slide 2

Slide 2 text

ࣗݾ঺հ .BLF-4* 73"3 ࢁాඒึʢ΍·ͩΈ΄ʣ ɾʮ*P5Λ΋ͬͱ਎ۙʹɺ΋ͬͱָ͘͠ɻະདྷ΁ʯ ɹɹͰ͓ͳ͡Έͷɺגࣜձࣾ$FOUFS2୅දऔక໾ ɾʮ*P5ओ්ʯͱ໊৐͍ͬͯ·͢ ɾঁࢠͩΒ͚ͷిࢠ޻࡞ओ࠵ ɾझຯɿϋϯζΦϯ΍ษڧձʹࢀՃ͢Δ͜ͱɻ ɹ

Slide 3

Slide 3 text

ΈΜͳͷࣗݾ঺հλΠϜ ̍ʣ໊͓લ ̎ʣԿΛ͍ͯ͠Δਓ͔ ̏ʣࠓ೔ͷϞνϕʔγϣϯ

Slide 4

Slide 4 text

ຊ೔ֶΜͰ͍͚ͨͩΔ͜ͱ ̍ʣRaspberryPiͷجૅ஌ࣝ ̎ʣσΟʔϓϥʔχϯάʹඞཁͳϥΠϒϥϦʔ౳ͷΠϯετʔϧखॱ ̏ʣσʔληοτMNISTΛ࢖ͬͨखॻ͖ೝࣝจࣈ ̐ʣRaspberryPiͰUSBΧϝϥΛ࢖͏खॱ ̑ʣֶशࡁΈϞσϧInceptionV3Λ࢖ͬͨը૾ೝࣝ ̒ʣը૾Λ࢖ͬͯσΟʔϓϥʔχϯά(CNN)ͤ͞Δϓϩηε

Slide 5

Slide 5 text

3BTQCFSSZ1Jͱ 64#ΧϝϥͰ͸͡ΊΑ͏ σΟʔϓϥʔχϯά

Slide 6

Slide 6 text

3BTQCFSSZ1JͷجૅΛֶͿ 3BTQCFSSZ1Jͷछྨ

Slide 7

Slide 7 text

3BTQCFSSZ1JͷجૅΛֶͿ 3BTQCFSSZ1Jͷछྨ ͋ͨΒ͘͠ ൃച͞Ε·ͨ͠ʂ ൢചՁ֨͸ ԁʙ

Slide 8

Slide 8 text

ٕదະऔಘػثΛ༻͍࣮ͨݧ౳ͷ ಛྫ੍౓ͷਃ੥͖ͯ͠·ͨ͠ʂʂ

Slide 9

Slide 9 text

ຊ೔ͷ࢖༻ػث 3BTQCFSSZ1JͷجૅΛֶͿ 3BTQCFSSZ1Jͷछྨ

Slide 10

Slide 10 text

ຊ೔࢖༻͢Δ3BTQCFSSZ1J.PEFM#͸ ɹ"3.$PSUFY"ʢίΞɺ()[ʣ ɹ #SPBEDPN#$. ετϨʔδɹNJDSP4% ɹ64#Y )%.*º -"/#"4&5 ɹΦʔσΟΦɹNNεςϨΦ ɹిݯɹ7ʗ"ʢ64#NJDSP#ʣ ɹେ͖͞ɹºNN ɹ8J'J͸()[ଳͷΈ࢖༻Ͱ͖·͢ɻ 3BTQCFSSZ1JͷجૅΛֶͿ 3BTQCFSSZ1J̏.PEFM#ͱ͸ ˞3BTQCFSSZ1J͸օ͞Μ͕࢖͍ͬͯΔ1$ΑΓ൓Ԡ଎౓͕ͱͯ΋஗͍Ͱ͢ɻ ɹʢ͙͢൓Ԡ͠ͳ͍͔ΒͱԿ౓΋ΫϦοΫͨ͠ΓίϚϯυΛ࠶౓ೖྗ͠ͳ͍ʣ

Slide 11

Slide 11 text

https://www.raspberrypi.org/ 3BTQCFSSZ1Jެࣜϖʔδ ֤छΠϯετʔϧ΍࢖͍ํͳͲϥζύΠͷ࠷৽৘ใ͸ͪ͜ΒΛνΣοΫ͠·͢ɻ 3BTQCFSSZ1JͷجૅΛֶͿ

Slide 12

Slide 12 text

https://www.raspberrypi.org/downloads/ ࠓճͷηϛφʔͰ͸ طʹ04͕Πϯετʔϧ ͞ΕͨNJDSP4% (# Λ ࢖͍·͢ɻ ͳͷͰΠϯετʔϧͷ ࡞ۀ͸ෆཁͰ͢ɻ ϥζύΠ͸3BTQCJBOʢϥζϏΞϯʣͱ͍͏-JOVYܥͷ04Λ࢖ͬͯಈ͔͠·͢ɻ /00#4 ψʔϒεʣͱ͍͏04ͷΠϯετʔϥʔΛ࢖͏ͱศརͰ͢ɻ 3BTQCFSSZ1Jͷ04ʹ͍ͭͯʢNJDSP4% 3BTQCFSSZ1JͷجૅΛֶͿ

Slide 13

Slide 13 text

3BTQCFSSZ1JͱσΟʔϓϥʔχϯάͰͲΜͳ͜ͱ͕Ͱ͖Δͷʁɹɹ ͖Ύ͏Γͷ඼࣭൑ผ ͔Β͋͛ͷ഑હ

Slide 14

Slide 14 text

ख़࿅ͨ͠ਓͷϊ΢ϋ΢ˠࣗಈԽ ˁ ͦΕͰ͸ɺ͓଴ͨͤ͠·ͨ͠ʂ ϥζύΠΛҰॹʹ४උ͍͖ͯ͠·͠ΐ͏ʂ Έͳ͞Μ͸ɺԿΛͭ͘Γ·͔͢ʁ

Slide 15

Slide 15 text

ຊ೔ͷ3BTQCFSSZ1Jपลػثߏ੒ͱ઀ଓ֬ೝ 3BTQCFSSZ1J̏.PEFM# 64#ΩʔϘʔυ 64#Ϛ΢ε 64#Χϝϥ )%.* σΟεϓϨΠ ిݯ͸·ͩೖΕͳ͍Ͱ σΟεϓϨΠ ɹిݯ NJDSP4% -"/έʔϒϧ

Slide 16

Slide 16 text

3BTQCFSSZ1JͷిݯΛೖΕͯىಈ ˞ͻͱੲલ3BTQCFSSZ1J͕ྲྀߦͬͨ࣌ͱ͸ҧͬͯࠓ͸ࣗಈͰϩάΠϯ͠·͢

Slide 17

Slide 17 text

˞ͷγεςϜͷจࣈ͕খͯ͘͞ݟʹ͍͘ͳ͋ʝͱ͍͏ਓ͸ઃఆมߋ 3BTQCFSSZ1Jͷจࣈʹ͍ͭͯ എܠը໘ΛӈΫϦοΫ

Slide 18

Slide 18 text

γεςϜͷจࣈ͕খ͍࣌͞͸എܠը໘ΛӈΫϦοΫ ʮσεΫτοϓͷઃఆʯΛબ୒ͨ͠Β"QQFBSBODF4FUUJOH͕ ։͘ͷͰɺʮ4ZTUFNʯͷ'POUͷϘλϯΛΫϦοΫ͢Δͱ ʮ1JDLB'POUʯ͕։͘ͷͰݟ΍͍͢4J[Fʹมߋ͍ͩ͘͞ɻ

Slide 19

Slide 19 text

ɹˢίίΛΫϦοΫ͢Δͱ ɹɹ-95FSNJOBM্ཱ͕͕ͪΔ -95FSNJOBMͷىಈ

Slide 20

Slide 20 text

λʔϛφϧͷจࣈ͕খ͍࣌͞͸ฤूΛΫϦοΫ ʮઃఆ 4 ʯΛબ୒ͨ͠Β৽͘͠΢Οϯυ΢͕։͘ͷͰɺ ʮελΠϧʯͷʰ୺຤ͷϑΥϯτʱͷϘλϯ෦෼ΛΫϦοΫ͢Δͱ ʮϑΥϯτͷબ୒ʯ͕։͘ͷͰݟ΍͍͢4J[Fʹมߋ͍ͩ͘͞ɻ

Slide 21

Slide 21 text

ࠓճ࢖༻͢ΔϓϩάϥϜݴޠ͸ʝ ✋QZUIPOΛ৮ͬͨ͜ͱ͕͋Δํ͸खΛ্͛ͯ ίʔυ͕γϯϓϧͰɺ૊ΈࠐΈ΍8FCΞϓϦͷ։ൃɺ·ͨਓ޻஌ೳ΍ɺ ϏοάσʔλղੳͳͲͰΑ͘࢖༻͞ΕΔ൚༻ϓϩάϥϛϯάݴޠͰ͢ɻ ʮ1ZUIPOʢύΠιϯʣͰ͢ʯ

Slide 22

Slide 22 text

ίϚϯυΛೖྗͯ͠Έ·͠ΐ͏ʂ ίϚϯυ໊ ˢˈϚʔΫ͸දࣔ͞Ε͍ͯΔͷͰɺˈ͸ೖΕΔඞཁ͕ͳ͍Ͱ͢ɻ -95FSNJOBMͷը໘ʹίϚϯυΛೖྗ͢Δ

Slide 23

Slide 23 text

5FOTPS'MPXͱ͸ 5FOTPS'MPXʢςϯιϧϑϩʔʣͱ͸ɺ͞·͟·ͳػցֶश ʢ.BDIJOF-FBSOJOHʣͷ෼໺Ͱ࢖༻͞Ε͍ͯΔ044ʢΦʔϓ ϯιϑτ΢ΣΞϥΠϒϥϦʣͰ͢ɻ(PPHMF͕:PV5VCF΍ (PPHMF຋༁ɺԻ੠ݕࡧͳͲͰ࢖༻͍ͯͨ͠πʔϧͰɺ 3BTQCFSSZ1J্Ͱ΋ಈ࡞͠·͢ɻ ,FSBTͱ͸ ,FSBTʢέϥεʣͱ͸ɺ1ZUIPOͰॻ͔Εͨχϡʔϥϧωοτ ϫʔΫͷϥΠϒϥϦͰ͢ɻ5FOTPS'MPXɺ$/5,ʢ.JDSPTPGU $PHOJUJWF5PPMLJUʣɺ5IFBOPʢςΞϊʣ্Ͱಈ࡞͢Δ͜ͱ͕ Ͱ͖·͢ɻࠓճ͸,FSBTΛ࢖ͬͯը૾ೝࣝΛ͓͜ͳ͍·͢ɻ IUUQTLFSBTJP ࠓճ࢖༻͢Δ̎ͭͷπʔϧ͸ʝ

Slide 24

Slide 24 text

σΟʔϓϥʔχϯάʹඞཁͳ ϥΠϒϥϦ౳ͷΠϯετʔϧखॱ طʹΠϯετʔϧࡁΈ ͳͷͰɺྲྀΕ͚ͩΛ આ໌͠·͢ɻ

Slide 25

Slide 25 text

ݱࡏͷ1ZUIPOͱQJQͷόʔδϣϯ֬ೝ ̍ʣ1ZUIPOʢ̏ܥʣͷόʔδϣϯΛ֬ೝ͢ΔίϚϯυ ̎ʣQJQ QZUIPOͷόʔδϣϯ؅ཧπʔϧʣ̏ܥͷόʔδϣϯΛ֬ೝ͢ΔίϚϯυ QJQGSPNVTSMJCQZUIPOEJTUQBDLBHFTQJQ(QZUIPO) QZUIPO7 QJQ7 1ZUIPO 1ZUIPOͷόʔδϣϯ͸̎ܥͱ ̏ܥ͕͋Γ·͢ɻࠓճ͸ܥΛ ѻ͍·͢ɻ7͚ͩେจࣈͰ͢ʂ QJQ͸ύοέʔδ؅ཧγεςϜͰ ͢ɻQJQ͸1ZUIPȌܥͷ؅ཧ

Slide 26

Slide 26 text

"5-"4ʢઢܗ୅਺ϥΠϒϥϦʣͷΠϯετʔϧ ɹɹ "VUPNBUJDBMMZ5VOFE-JOFBS"MHFCSB4PGUXBSFʣ TVEPBQUJOTUBMMMJCBUMBTCBTFEFW "5-"4ʢઢܗ୅਺ϥΠϒϥϦʣͷΠϯετʔϧखॱ ్தͰΠϯετʔϧ͕ࢭ·ͬͯ ଓߦ͠·͔͢ʁ<:O>ͱฉ͔Ε·͢ͷͰ :Λೖྗޙ&OUFSͯ͠ ଓ͚͍ͯͩ͘͞ɻ ஫ҙɿ طʹΠϯετʔϧࡁΈͰ͢ ίϚϯυΛೖྗ͢Δඞཁ͸͋ Γ·ͤΜɻ

Slide 27

Slide 27 text

ȊQZͷΠϯετʔϧखॱ IQZ ΤΠνϑΝΠϒύΠ ͷΠϯετʔϧ QZUIPOͰ )%' )JFSBSDIJDBM %BUB'PSNBU ܗࣜΛ ѻ͑ΔΑ͏ʹ͠·͢ɻ TVEPBQUHFUJOTUBMMQZUIPOIQZ ్தͰΠϯετʔϧ͕ࢭ·ͬͯ ଓߦ͠·͔͢ʁ<:O> ͱฉ͔Ε·͢ͷͰ :Λೖྗޙ&OUFSͯ͠ ଓ͚͍ͯͩ͘͞ɻ ˞LFSBTͳͲͷχϡʔϥϧωοτϫʔΫϥΠϒϥϦͷֶश݁ՌΛอଘ͢ΔϑΥʔϚοτ ஫ҙɿ طʹΠϯετʔϧࡁΈͰ͢ ίϚϯυΛೖྗ͢Δඞཁ͸͋ Γ·ͤΜɻ

Slide 28

Slide 28 text

5FOTPSqPXͷΠϯετʔϧखॱ 5FOTPSqPXͷΠϯετʔϧ TVEPQJQJOTUBMMUFOTPSqPX ʢ͜͜·Ͱ෼ఔ౓ʣ 3VOOJOHTFUVQQZCEJTU@XIFFMGPSOVNQZʜ ʢ͜ͷ͋ͱ͞Βʹ̎̌෼ఔ౓͔͔Γ·͢ɻʣ ஫ҙɿ طʹΠϯετʔϧࡁΈͰ͢ ίϚϯυΛೖྗ͢Δඞཁ͸͋ Γ·ͤΜɻ

Slide 29

Slide 29 text

,FSBTͷΠϯετʔϧखॱ ,FSBTͷΠϯετʔϧ TVEPQJQJOTUBMMLFSBT ஫ҙɿ طʹΠϯετʔϧࡁΈͰ͢ ίϚϯυΛೖྗ͢Δඞཁ͸͋ Γ·ͤΜɻ

Slide 30

Slide 30 text

0QFO$7ͷΠϯετʔϧखॱ 0QFO$7ͱ͸ TVEPBQUHFUJOTUBMMQZUIPOPQFODW ஫ҙɿ طʹΠϯετʔϧࡁΈͰ͢ ίϚϯυΛೖྗ͢Δඞཁ͸͋ Γ·ͤΜɻ ͜ΕͰඞཁͳΠϯετʔϧ͕׬ྃ͠·ͨ͠ʂ ը૾΍ಈըΛॲཧ͢Δػೳ͕࣮૷͞Ε͍ͯΔϥΠϒϥϦ

Slide 31

Slide 31 text

͜ΕͰUFOTPSqPXͱ,FSBTΛ࢖͏ ؀ڥ͕੔͍·ͨ͠ʂʂ ͜ΕͰ3BTQCFSSZ1JͰσΟʔϓϥʔχϯά͕ࢼͤ·͢ʂ

Slide 32

Slide 32 text

ਓ޻஌ೳͱػցֶशͱਂ૚ֶशͷؔ܎ ਓ޻஌ೳ ػցֶश ਂ૚ ֶश ೖྗ ग़ྗ ʢػցʹίϯϐϡʔλʹֶशͤ͞Δ͜ͱʣ ʢόΠΦςΫϊϩδʔ ɹɹͳͲ΋͋Δʣ

Slide 33

Slide 33 text

σΟʔϓϥʔχϯάʢਂ૚ֶशʣͱ͸ σΟʔϓϥʔχϯάͱ͸Ի੠ͷೝࣝɺը૾ೝࣝɺҟৗݕ஌ͳͲ͢ΔͨΊʹɺίϯϐϡʔ λʔʹֶशͤ͞Δͻͱͭͷख๏Ͱ͢ɻਓؒͷ೴ͷதʹ͋ΔχϡʔϩϯΛਅࣅͨχϡʔϥϧ ωοτϫʔΫͱݺ͹ΕΔख๏ͷ֊૚ΛਂΊͨΞϧΰϦζϜʹͳΓ·͢ɻ தؒ૚ͷ֊૚͕ΑΓਂ͘ͳ͍ͬͯΔ ਓؒͷ೴ͷதʹ͋Δχϡʔϩϯ ܗࣜχϡʔϩϯ В y x1 x2 x3 xN w1 w2 w3 wN …………… …… σΟʔϓϥʔχϯά ೖྗ૚தؒ૚ɹग़ྗ૚ χϡʔϥϧωοτϫʔΫ ೖྗ૚தؒ૚ɹग़ྗ૚

Slide 34

Slide 34 text

ը૾ೝࣝͤ͞Δҝͷ̎ͭͷϑΣʔζ ࣗಈͰը૾ೝࣝΛͤ͞Δʹ͸େ͖͘ೋͭͷϑΣʔζ͕͋Γ·͢ɻ ୈ̍ϑΣʔζɿֶशϞσϧΛͭ͘Δ ୈ̎ϑΣʔζɿֶशࡁΈϞσϧΛ࢖͏ σʔλ ॲཧ σʔλ ऩू ػցֶश ਂ૚ֶश ֶशͤ͞Δը૾ ֶशࡁΈ Ϟσϧ ֶश༻ σʔληοτ ೝ͍ࣝͤͨ͞ը૾ ೖྗ ग़ྗ ֶशࡁΈ Ϟσϧ ਪ࿦ ը૾ೝࣝΛ ࢖ͬͨॲཧ

Slide 35

Slide 35 text

·ͣ͸ֶशϞσϧΛͭ͘ΔϑΣʔζΛֶ΅͏ʂ ࣗಈͰը૾ೝࣝΛͤ͞Δʹ͸େ͖͘ೋͭͷϑΣʔζ͕͋Γ·͢ɻ ୈ̍ϑΣʔζɿֶशϞσϧΛͭ͘Δ ୈ̎ϑΣʔζɿֶशࡁΈϞσϧΛ࢖͏ σʔλ ॲཧ σʔλ ऩू ػցֶश ਂ૚ֶश ֶशͤ͞Δը૾ ֶशࡁΈ Ϟσϧ ֶश༻ σʔληοτ ೝ͍ࣝͤͨ͞ը૾ ೖྗ ग़ྗ ֶशࡁΈ Ϟσϧ ਪ࿦ ը૾ೝࣝΛ ࢖ͬͨॲཧ

Slide 36

Slide 36 text

·ͣ͸ֶशϞσϧΛͭ͘ΔϑΣʔζΛֶ΅͏ʂ ਓखෆ଍Ͱ ༣ศ෺Λ࢓෼͚Δ ͷ͕ େมͰ͢ɻ ͋ͳͨͳΒɺͲ͏΍ͬͯ༣ศہͷനϠΪ͞ΜΛॿ͚·͔͢ʁʁ ༣ศ൪߸Ͱ ࢓෼͚͍ͯ·͢ɻ

Slide 37

Slide 37 text

·ͣ͸ֶशϞσϧΛͭ͘ΔϑΣʔζΛֶ΅͏ʂ खॻ͖༣ศ൪߸ͷ ਺ࣈΛࣗಈͰೝࣝ ͤ͞Δͱָͩͳʂ ·ͣ͸͍ΖΜͳਓͷ खॻ͖਺ࣈΛ ूΊΑ͏ʂ ͏Θʙ࢓෼͚Δ ྔ͕ଟͯ͘େม ͩ͠ɺखॻ͖਺ ࣈͬͯಡΈͮΒ ͍ͳ͋ɻ खॻ͖਺ࣈʢ༣ศ൪߸ʣΛը૾ೝࣝͰ͖ΔΑ͏ʹ͠Α͏ʂ

Slide 38

Slide 38 text

·ͣ͸ֶशϞσϧΛͭ͘ΔϑΣʔζΛֶ΅͏ʂ ֶशϞσϧ σʔληοτ ɹ࡞੒ ֶश༻σʔλɹɹɹɹɹɹςετ༻σʔλ ूΊ͖ͯͨखॻ͖਺ࣈσʔλ ूΊͨσʔλ͔ΒσʔληοτΛ࡞੒ֶͯ͠शϞσϧΛ࡞Δ खॻ͖਺ࣈʢ༣ศ൪߸ʣΛը૾ೝࣝͰ͖ΔΑ͏ʹ͠Α͏ʂ ֶश σʔλॲཧ

Slide 39

Slide 39 text

·ͣ͸ֶशϞσϧΛͭ͘ΔϑΣʔζΛֶ΅͏ʂ ֶश༻σʔλɹɹɹɹɹɹςετ༻σʔλ ूΊ͖ͯͨखॻ͖਺ࣈσʔλ ֶशϞσϧ ςετ༻σʔλͰ ਖ਼౴཰Λௐ΂Δ ϓϩάϥϛϯά ςετ ೲ඼ ׬ ੒ ૊ΈࠐΈ

Slide 40

Slide 40 text

खॻ͖਺ࣈΛը૾ೝࣝͤ͞ΔͨΊͷ޻ఔ ୈ̍ϑΣʔζɿֶशϞσϧΛͭ͘Δ σʔλ ॲཧ σʔλ ऩू ػցֶश ਂ૚ֶश ֶशͤ͞Δը૾ ֶशࡁΈ Ϟσϧ ֶश༻ σʔληοτ ֶशϞσϧ

Slide 41

Slide 41 text

खॻ͖਺ࣈͷσʔληοτ./*45ʢΤϜχετʣ ػցֶशΛֶͿਓʹͱͬͯ௒༗໊ͳσʔληοτͰ͢ɻ͔Β· Ͱͷສݸͷखॻ͖਺ࣈʢֶश༻̒ສݸɾςετ༻̍ສݸʣͷը૾ ͱͦΕʹରԠ͢Δ౴͑ʢϥϕϧʣ͕ఏڙ͞Ε͍ͯ·͢ɻ

Slide 42

Slide 42 text

./*45ͷը૾ɹɹɹɹը૾Λd·Ͱͷ਺ࣈͰදݱͨ͠ߦྻ

Slide 43

Slide 43 text

5IPOOZ 1ZUIPO*%&1ZUIPOͷॳ৺ऀ༻౷߹؀ڥ Λ࢖͏ UIPOOZͱ͍͏σϑΥϧτͰ ೖ͍ͬͯΔQZUIPOͷ*%&Λ ࠓճ࢖͍·͢ɻ ./*45ΛಡΈࠐΉϓϩάϥϜ NOJTUQZΛ ࣄલʹೖΕ͓͖ͯ·ͨ͠ͷͰ ։͖·͢ɻ UIPOOZNOJTUQZ ͱλʔϛφϧͰଧ͍ͬͯͩ͘͞ɻ खॻ͖਺ࣈͷσʔληοτ./*45

Slide 44

Slide 44 text

from keras.datasets import mnist import matplotlib.pyplot as plt (X_train, y_train), (X_test, y_test) = mnist.load_data() print("label",y_train[0]) plt.imshow(X_train[0].reshape(28,28),cmap='Greys') plt.show() खॻ͖਺ࣈͷσʔληοτ./*45Λ࢖͓͏

Slide 45

Slide 45 text

Slide 46

Slide 46 text

Slide 47

Slide 47 text

Slide 48

Slide 48 text

5IPOOZ 1ZUIPO*%&1ZUIPOͷॳ৺ऀ༻౷߹؀ڥ Λ࢖͏ 3VOϘλϯΛԡ࣮ͯ͠ߦͤ͞Α͏ʂ खॻ͖਺ࣈͷσʔληοτ./*45 &Run mnist1.py Using TensorFlow backend. WARNING: Logging before flag parsing goes to stderr. W0823 09:50:29.745067 1995954896 deprecation_wrapper.py:118] From /home/pi/.local/lib/python3.7/site- packages/tensorflow/__init__.py:98: The name tf.AUTO_REUSE is deprecated. Please use tf.compat.v1.AUTO_REUSE instead. WARNING͕͍͔ͭ͘ग़·͕͢ ໰୊ͳ͍ͷͰ͠͹Β͘଴ͪ·͢

Slide 49

Slide 49 text

Slide 50

Slide 50 text

खॻ͖਺ࣈͷσʔληοτ./*45Λ࢖ͬͯ ֶशϞσϧΛ࡞Ζ͏ʂ ./*45Λ࢖ֶͬͯशϞσϧΛ ࡞ΔϓϩάϥϜ NOJTUQZΛ ࣄલʹೖΕ͓͖ͯ·ͨ͠ͷͰ 5IPOOZͷ-PBEΛΫϦοΫͯ͠ ։͖NOJTUQZΛબ୒ͯ͠ ։͖·͢ɻ

Slide 51

Slide 51 text

खॻ͖਺ࣈͷσʔληοτ./*45Λ࢖ͬͯ ֶशϞσϧΛ࡞Ζ͏ʂ from keras.datasets import mnist from keras.models import Sequential from keras.layers.core import Dense, Activation from keras.utils import np_utils (X_train, y_train), (X_test, y_test) = mnist.load_data() σʔληοτͷಡΈࠐΈ 9@USBJOֶशͤ͞Δը૾σʔλ Z@USBJOֶशͤ͞Δϥϕϧσʔλ 9@UFTUֶशͤͨ͞ϞσϧΛςετ͢Δςετ༻ը૾σʔλ Z@UFTUֶशͤͨ͞ϞσϧΛςετ͢Δςετ༻ϥϕϧσʔλ NOJTUͷΠϯϙʔτ

Slide 52

Slide 52 text

खॻ͖਺ࣈͷσʔληοτ./*45Λ࢖ͬͯ ֶशϞσϧΛ࡞Ζ͏ʂ 9@USBJO9@USBJOSFTIBQF 9@UFTU9@UFTUSFTIBQF ֶशͤ͞Δը૾σʔλͱςετ༻ը૾σʔλΛ ϦγΣΠϓ͠·͢ɻ ./*45ͷը૾σʔλ ɹɹɹɹɹɹϐΫηϧºϐΫηϧϐΫηϧ ̍ϐΫηϧͷάϨʔεέʔϧ஋Λ̌ʙ̎̑̑ͷ਺ࣈͰදݱ ɹɹͰׂΔ͜ͱͰ̍ϐΫηϧͷ஋Λ̌ʙ̍ͷؒʹ͠·͢ ʢਖ਼نԽɿ̌ʙ̍·Ͱ͔͠ѻ͑ͳ͍ͷͰ߹Θͤ·͢ʣ

Slide 53

Slide 53 text

./*45ͷը૾ɹɹɹɹը૾Λd·Ͱͷ਺ࣈͰදݱͨ͠ߦྻ

Slide 54

Slide 54 text

खॻ͖਺ࣈͷσʔληοτ./*45Λ࢖ͬͯ ֶशϞσϧΛ࡞Ζ͏ʂ Z@USBJOOQ@VUJMTUP@DBUFHPSJDBM Z@USBJO Z@UFTUOQ@VUJMTUP@DBUFHPSJDBM Z@UFTU ,FSBT͸ϥϕϧΛόΠφϦʔͰ͔͠࢖͑ͳ͍ͷͰɺ ϥϕϧΛ̌ͱ̍ͷߦྻʹ͢ΔॲཧͰ͢

Slide 55

Slide 55 text

खॻ͖਺ࣈͷσʔληοτ./*45Λ࢖ͬͯ ֶशϞσϧΛ࡞Ζ͏ʂ NPEFM4FRVFOUJBM < %FOTF JOQVU@TIBQF "DUJWBUJPO bTJHNPJE %FOTF "DUJWBUJPO TPGUNBY > ֶशϞσϧΛ࡞͍ͬͯ·͢ɻ

Slide 56

Slide 56 text

खॻ͖਺ࣈͷσʔληοτ./*45Λ࢖ͬͯ ֶशϞσϧΛ࡞Ζ͏ʂ NPEFMDPNQJMF MPTTDBUFHPSJDBM@DSPTTFOUSPQZ PQUJNJ[FSTHE NFUSJDT<BDDVSBDZ> NPEFMpU 9@USBJO Z@USBJO CBUDI@TJ[F WFSCPTF FQPDIT WBMJEBUJPO@TQMJU ֶशϞσϧΛ࡞͍ͬͯ·͢ɻ

Slide 57

Slide 57 text

खॻ͖਺ࣈͷσʔληοτ./*45Λ࢖ͬͨ ֶशϞσϧͷਖ਼౴཰Λग़ྗ TDPSFNPEFMFWBMVBUF 9@UFTU Z@UFTU WFSCPTF QSJOU UFTUBDDVSBDZ TDPSF<> είΞʢਖ਼౴཰ʣΛग़ྗ͠·͢

Slide 58

Slide 58 text

5IPOOZ 1ZUIPO*%&1ZUIPOͷॳ৺ऀ༻౷߹؀ڥ Λ࢖͏ 3VOϘλϯΛԡ࣮ͯ͠ߦͤ͞Α͏ʂ खॻ͖਺ࣈͷσʔληοτ./*45 &Run mnist2.py Using TensorFlow backend. WARNING: Logging before flag parsing goes to stderr. W0823 09:50:29.745067 1995954896 deprecation_wrapper.py:118] From /home/pi/.local/lib/python3.7/site- packages/tensorflow/__init__.py:98: The name tf.AUTO_REUSE is deprecated. Please use tf.compat.v1.AUTO_REUSE instead. WARNING͕͍͔ͭ͘ग़·͕͢ ໰୊ͳ͍ͷͰ͠͹Β͘଴ͪ·͢

Slide 59

Slide 59 text

5IPOOZ 1ZUIPO*%&1ZUIPOͷॳ৺ऀ༻౷߹؀ڥ Λ࢖͏ खॻ͖਺ࣈͷσʔληοτ./*45 ਖ਼౴཰͸75.27%

Slide 60

Slide 60 text

վળͯ͠ਖ਼౴཰Λ͋͛Α͏ʂ

Slide 61

Slide 61 text

վળͯ͠ਖ਼౴཰Λ͋͛Α͏ʂ NPEFM4FRVFOUJBM < %FOTF JOQVU@TIBQF "DUJWBUJPO bTJHNPJE %FOTF "DUJWBUJPO TPGUNBY > TJHNPJEΛ SFMVʹมߋ͢Δ

Slide 62

Slide 62 text

Slide 63

Slide 63 text

NPEFMDPNQJMF MPTTDBUFHPSJDBM@DSPTTFOUSPQZ PQUJNJ[FSTHE NFUSJDT<BDDVSBDZ> NPEFMpU 9@USBJO Z@USBJO CBUDI@TJ[F WFSCPTF FQPDIT WBMJEBUJPO@TQMJU FQPDIT͔ΒFQPDIT ʹมߋ͠Α͏ վળͯ͠ਖ਼౴཰Λ͋͛Α͏ʂ

Slide 64

Slide 64 text

࣮ࡍʹखͰॻ͍ͨ਺ࣈΛಡΈࠐΜͰ ./*45Λ࢖ͬͯ൑ผ͠Α͏ʂ

Slide 65

Slide 65 text

JNQPSUOVNQZBTOQ JNQPSUDW JNQPSUNBUQMPUMJCQZQMPUBTQMU GSPNLFSBTEBUBTFUTJNQPSUNOJTU GSPNLFSBTNPEFMTJNQPSU4FRVFOUJBM GSPNLFSBTMBZFSTDPSFJNQPSU%FOTF "DUJWBUJPO GSPNLFSBTVUJMTJNQPSUOQ@VUJMT GSPNLFSBTNPEFMTJNQPSUMPBE@NPEFM ൑ผʹඞཁͳ΋ͷΛΠϯϙʔτ͠·͢ ࣮ࡍʹखͰॻ͍ͨ਺ࣈΛಡΈࠐΜͰ ./*45Λ࢖ͬͯ൑ผ͠Α͏ʂ

Slide 66

Slide 66 text

9@USBJO Z@USBJO 9@UFTU Z@UFTU NOJTUMPBE@EBUB 9@USBJO9@USBJOSFTIBQF 9@UFTU9@UFTUSFTIBQF Z@USBJOOQ@VUJMTUP@DBUFHPSJDBM Z@USBJO Z@UFTUOQ@VUJMTUP@DBUFHPSJDBM Z@UFTU NPEFM4FRVFOUJBM < %FOTF JOQVU@TIBQF "DUJWBUJPO SFMV %FOTF "DUJWBUJPO TPGUNBY > NPEFMDPNQJMF MPTTDBUFHPSJDBM@DSPTTFOUSPQZ PQUJNJ[FSTHE NFUSJDT<BDDVSBDZ> NPEFMpU 9@USBJO Z@USBJO CBUDI@TJ[F WFSCPTF FQPDIT WBMJEBUJPO@TQMJU TDPSFNPEFMFWBMVBUF 9@UFTU Z@UFTU WFSCPTF QSJOU UFTUBDDVSBDZ TDPSF<> ઌ΄ͲͷNOJTUQZͱಉ͡ॲཧͰ ਖ਼౴཰Λग़͠·͢ɻ ʢ࣌ؒͷؔ܎ͰFQPDITʹͱ͠·͢ʣ ࣮ࡍʹखͰॻ͍ͨ਺ࣈΛಡΈࠐΜͰ ./*45Λ࢖ͬͯ൑ผ͠Α͏ʂ

Slide 67

Slide 67 text

NPEFMTBWF ./*45I JNHDWJNSFBE KQH ઌ΄ͲͷNOJTUQZͱಉ͡ॲཧͰ ࡞ֶͬͨशϞσϧΛอଘ͠·͢ɻ ࠓճ͸ࢲ͕ॻ͍ͨ਺ࣈͷ̓ͷը૾Λ ೖΕ͓͍ͯͨͷͰɺ ͦΕΛ൑ผͯ͠Έ·͠ΐ͏ʂ ࣮ࡍʹखͰॻ͍ͨ਺ࣈΛಡΈࠐΜͰ ./*45Λ࢖ͬͯ൑ผ͠Α͏ʂ KQH ʻʻ͜ͷը૾͸ ɹɹJ1IPOFͷϝϞͰॻ͍ͯɺΩϟϓνϟʔͨ͠ ɹɹࣸਅͷαΠζΛฤूͰτϦϛϯάͨ͠΋ͷͰ͢ɻ

Slide 68

Slide 68 text

HSBZDWDWU$PMPS JNH DW$0-03@#(3(3": DWJNXSJUF HSBZQOH HSBZ PQFODWΛ࢖ͬͯը૾ΛάϨʔεέʔϧʹ͠·͢ ࣮ࡍʹखͰॻ͍ͨ਺ࣈΛಡΈࠐΜͰ ./*45Λ࢖ͬͯ൑ผ͠Α͏ʂ KQH HSBZQOH

Slide 69

Slide 69 text

@ CJOBSZDWUISFTIPME HSBZ DW5)3&4)@#*/"3: DWJNXSJUF CJOBSZQOH CJOBSZ PQFODWΛ࢖ͬͯը૾Λ̎஋Խ͠·͢ ࣮ࡍʹखͰॻ͍ͨ਺ࣈΛಡΈࠐΜͰ ./*45Λ࢖ͬͯ൑ผ͠Α͏ʂ CJOBSZQOH HSBZQOH

Slide 70

Slide 70 text

OFHBQPTJDWCJUXJTF@OPU CJOBSZ DWJNXSJUF OFHBQPTJQOH OFHBQPTJ PQFODWΛ࢖ͬͯը૾ͷനࠇΛ൓స͠·͢ ࣮ࡍʹखͰॻ͍ͨ਺ࣈΛಡΈࠐΜͰ ./*45Λ࢖ͬͯ൑ผ͠Α͏ʂ CJOBSZQOH OFHBQPTJQOH

Slide 71

Slide 71 text

CMVSDW(BVTTJBO#MVS OFHBQPTJ DWJNXSJUF CMVSQOH CMVS PQFODWΛ࢖ͬͯը૾ʹ΅͔͠Λ͍ΕΔ ࣮ࡍʹखͰॻ͍ͨ਺ࣈΛಡΈࠐΜͰ ./*45Λ࢖ͬͯ൑ผ͠Α͏ʂ OFHBQPTJQOH CMVSQOH

Slide 72

Slide 72 text

PQFODWΛ࢖ͬͯը૾αΠζΛºʹมߋ͢Δ ࣮ࡍʹखͰॻ͍ͨ਺ࣈΛಡΈࠐΜͰ ./*45Λ࢖ͬͯ൑ผ͠Α͏ʂ CMVSQOH JNHDWSFTJ[F CMVS DW*/5&3@$6#*$

Slide 73

Slide 73 text

ը૾αΠζΛºʹมߋͨ͠΋ͷΛલॲཧͯ͠ ઌ΄Ͳ࡞ֶͬͨशϞσϧΛ࢖ͬͯ൑ผ͠ Ұ൪Մೳੑ͕ߴ͍΋ͷΛදࣔ͢Δɻ ࣮ࡍʹखͰॻ͍ͨ਺ࣈΛಡΈࠐΜͰ ./*45Λ࢖ͬͯ൑ผ͠Α͏ʂ <> 9BQQFOE JNH 9OQBTBSSBZ 9 99 99SFTIBQF MFO 9 SFTVMUNPEFMQSFEJDU 9 QSJOU SFTVMUBSHNBY

Slide 74

Slide 74 text

3VOϘλϯΛԡ࣮ͯ͠ߦͤ͞Α͏ʂ &Run mnist2.py Using TensorFlow backend. WARNING: Logging before flag parsing goes to stderr. W0823 09:50:29.745067 1995954896 deprecation_wrapper.py:118] From /home/pi/.local/lib/python3.7/site- packages/tensorflow/__init__.py:98: The name tf.AUTO_REUSE is deprecated. Please use tf.compat.v1.AUTO_REUSE instead. WARNING͕͍͔ͭ͘ग़·͕͢ ໰୊ͳ͍ͷͰ͠͹Β͘଴ͪ·͢ ࣮ࡍʹखͰॻ͍ͨ਺ࣈΛಡΈࠐΜͰ ./*45Λ࢖ͬͯ൑ผ͠Α͏ʂ

Slide 75

Slide 75 text

3BTQCFSSZ1JͰ64#ΧϝϥΛ࢖͏खॱ ࣗ෼ͷखॻ͖਺ࣈΛ൑ผ͠Α͏ʂ

Slide 76

Slide 76 text

3BTQCFSSZ1JͷઃఆΛมߋ͢Δ ϝχϡʔͷɹɹΛΫϦοΫͯ͠ ઃఆʼ3TQCFSSZ1Jͷઃఆ ΛΫϦοΫ͠·͢ɻ

Slide 77

Slide 77 text

3BTQCFSSZ1JͷઃఆΛมߋ͢Δ ֤ࣗͷύεϫʔυ ֤ࣗͷύεϫʔυ ɹ ⁞3BTQCFSSZ1Jͷઃఆ ɹ্ཱ͕͕ͪͬͨΒ ɹʮΠϯλʔϑΣΠεʯΛ ɹɹબ୒͍ͯͩ͘͠͞ɻ

Slide 78

Slide 78 text

3BTQCFSSZ1JͷઃఆΛมߋ͢Δ ɹΧϝϥΛ༗ޮʹͯ͠ ɹͯ͠0,Λԡͨ͠Β ɹ࠶ىಈ͠·͢ɻ

Slide 79

Slide 79 text

64#Χϝϥͷಈ࡞ςετʢHVWDWJFXΛ࢖͍·͢ʣ TVEPBQUHFUJOTUBMMHVWDWJFX ͱೖྗͯ͠Πϯετʔϧ͢Δ ్தͰΠϯετʔϧ͕ࢭ·ͬͯ ଓߦ͠·͔͢ʁ<:O> ͱฉ͔Ε·͢ͷͰ:Λೖྗޙ&OUFSͯ͠ ଓ͚͍ͯͩ͘͞ɻ

Slide 80

Slide 80 text

Ұ୴Πϯετʔϧ͢Δͱϝχϡʔ͔Β΋ىಈͰ͖·͢ɻ ϝχϡʔͷɹɹΛΫϦοΫͯ͠ α΢ϯυͱϏσΦʼHVWDWJFX ΛΫϦοΫ͠·͢ɻ

Slide 81

Slide 81 text

64#Χϝϥͷಈ࡞ςετʢHVWDWJFXΛ࢖͍·͢ʣ ΩʔϘʔυͷ*Λԡ͢ͱ Ωϟϓνϟʔ͕ࡱΕ·͢ ϝχϡʔ͔ΒͰͳ͘ λʔϛφϧ͔Β HVWDWJFX ͱೖྗͯ͠΋0,Ͱ ͢ɻ

Slide 82

Slide 82 text

64#Χϝϥͷಈ࡞ςετʢHVWDWJFXΛ࢖͍·͢ʣ ίϯτϥε τΛௐ੔͢ Δɻ ΩʔϘʔυ ͷ*Λԡ͢ͱ Ωϟϓ νϟʔ͕ࡱ Ε·͢

Slide 83

Slide 83 text

64#Χϝϥͷಈ࡞ςετʢࡱͬͨࣸਅΛ֬ೝ͠·͠ΐ͏ʣ อଘ͞ΕͨࣸਅΛӈΫϦοΫ͠ ͯ ϑΝΠϧ໊ͷมߋΛԡͯ͠ ੩ࢭըͷ໊લΛίϐʔ͠·͠ΐ ͏ ֬ೝ͠·͠ΐ͏ɻ ʢNZ@QIPUP൪߸KQHͰ ɹอଘ͞Ε·͢ɻʣ ϑΥϧμϚʔΫΛΫϦοΫͯ͠ QJϑΥϧμΛ։͖·͠ΐ͏

Slide 84

Slide 84 text

ֶशࡁΈϞσϧΛ࢖ͬͨը૾ೝࣝ

Slide 85

Slide 85 text

Slide 86

Slide 86 text

ֶशࡁΈϞσϧΛ࢖ͬͯΈΑ͏ʂ ୈ̍ϑΣʔζɿֶशϞσϧΛͭ͘Δ ୈ̎ϑΣʔζɿֶशࡁΈϞσϧΛ࢖͏ σʔλ ॲཧ σʔλ ऩू ػցֶश ਂ૚ֶश ֶशͤ͞Δը૾ ֶशࡁΈ Ϟσϧ ֶश༻ σʔληοτ ೝ͍ࣝͤͨ͞ը૾ ೖྗ ग़ྗ ֶशࡁΈ Ϟσϧ ਪ࿦ ը૾ೝࣝΛ ࢖ͬͨॲཧ

Slide 87

Slide 87 text

ֶशࡁΈϞσϧ*ODFQUJPO7Λ࢖ͬͨը૾ೝࣝʹ௅ઓ *ODFQUJPO7ͱ͸ (PPHMF͕։ൃͨ͠σΟʔϓϥʔ χϯάΛ࢖ֶͬͨशϞσϧ σʔληοτ*."(&/&5Ͱֶश ͞Ε͍ͯ·͢ɻ 5FOTPSqPXͰར༻ग़དྷΔ༷ʹҰ ൠެ։͞Ε͍ͯ·͢ɻ

Slide 88

Slide 88 text

ݩʹͳΔը૾ɹɹ໿ສຕ ϥϕϧͷΧςΰϦ໿ສઍछྨ ͷσʔληοτ ֶशࡁΈϞσϧ*ODFQUJPO7Λ࢖ͬͨը૾ೝࣝʹ௅ઓ ֶश༻σʔληοτ*NBHF/FUͱ͸͜ͷ෦෼

Slide 89

Slide 89 text

5IPOOZͷ-PBEΛΫϦοΫͯ͠ ։͖NPEFMQZΛબ୒ͯ͠ ։͖·͢ɻ ֶशࡁΈϞσϧΛ࢖ͬͯը૾ೝࣝ͢Δ

Slide 90

Slide 90 text

from keras.preprocessing import image from keras.applications.inception_v3 \ import preprocess_input, decode_predictions, InceptionV3 import numpy as np model = InceptionV3(weights='imagenet') img_path=' banana.jpg' img = image.load_img(img_path, target_size=(299,299)) x = image.img_to_array(img) x = np.expand_dims(x, axis=0) x = preprocess_input(x) preds = model.predict(x) ֶशࡁΈϞσϧΛ࢖ͬͯը૾ೝࣝ͢Δ޻ఔͱ͸ CBOBOBKQH ɹɹɹˣ *ODFQUJPO7ͱ ͍͏ֶशࡁΈ ϞσϧΛ࢖ͬͯ ը૾ೝࣝ ɹɹˣ είΞͰ݁Ռ͕ ͰΔ

Slide 91

Slide 91 text

όφφը૾Ͱ࣮ߦ ̍ʣϝχϡʔͷ36/ϘλϯΛԡ͢ ̎ʣ4IFMMͷʮ3VONPEFMQZʯͱදࣔ͞ΕͯϓϩάϥϜ͕࣮ߦ͞Ε·͢ɻ 8BSOJOH͕ग़·͕͢໰୊ͳ͍Ͱ͢ɻ෼ඵఔ౓Ͱ݁Ռ͕ग़·͢ɻ ̍౓͚ͩΫϦοΫͯ͠ ͍ͩ͘͞ɻԿ౓΋ԡ͞ ͳ͍Α͏ʹɻʢॲཧ͕ ஗͘ͳΓ·͢ɻ

Slide 92

Slide 92 text

from keras.preprocessing import image from keras.applications.inception_v3 \ import preprocess_input, decode_predictions, InceptionV3 import numpy as np model = InceptionV3(weights='imagenet') img_path=' banana.jpg' img = image.load_img(img_path, target_size=(299,299)) x = image.img_to_array(img) x = np.expand_dims(x, axis=0) x = preprocess_input(x) preds = model.predict(x) print('Predicted:') for p in decode_predictions(preds, top=5)[0]: print("Score {}, Label {}".format(p[2],p[1])) ֶशࡁΈϞσϧΛ࢖ͬͨॲཧͷઆ໌ɿͦͷ̍ ը૾ͷಡΈࠐΈ΍ ֶशࡁΈϞσϧͷ *ODFQUJPO7ͷಡΈ ࠐΈɺ1ZUIPOֶज़ ܭࢉϥΠϒϥϦʔ ͷOVNQZΛΠϯ ϙʔτ͢Δ

Slide 93

Slide 93 text

from keras.preprocessing import image from keras.applications.inception_v3 \ import preprocess_input, decode_predictions, InceptionV3 import numpy as np model = InceptionV3(weights='imagenet') img_path=' banana.jpg' img = image.load_img(img_path, target_size=(299,299)) x = image.img_to_array(img) x = np.expand_dims(x, axis=0) x = preprocess_input(x) preds = model.predict(x) print('Predicted:') for p in decode_predictions(preds, top=5)[0]: print("Score {}, Label {}".format(p[2],p[1])) ֶशࡁΈϞσϧΛ࢖ͬͨॲཧͷઆ໌ɿͦͷ *NBHF/FUͰֶश͠ ͨॏΈͱ *ODFQUJPO7 ͷجຊϞσϧ͕ಡ Έࠐ·ΕΔ

Slide 94

Slide 94 text

from keras.preprocessing import image from keras.applications.inception_v3 \ import preprocess_input, decode_predictions, InceptionV3 import numpy as np model = InceptionV3(weights='imagenet') img_path=' banana.jpg' img = image.load_img(img_path, target_size=(299,299)) x = image.img_to_array(img) x = np.expand_dims(x, axis=0) x = preprocess_input(x) preds = model.predict(x) print('Predicted:') for p in decode_predictions(preds, top=5)[0]: print("Score {}, Label {}".format(p[2],p[1])) ֶशࡁΈϞσϧΛ࢖ͬͨॲཧͷઆ໌ɿͦͷ ೝ͍ࣝͤͨ͞ը૾ ͷࢦఆͱը૾ͷα Πζͷॲཧ

Slide 95

Slide 95 text

from keras.preprocessing import image from keras.applications.inception_v3 \ import preprocess_input, decode_predictions, InceptionV3 import numpy as np model = InceptionV3(weights='imagenet') img_path=' banana.jpg' img = image.load_img(img_path, target_size=(299,299)) x = image.img_to_array(img) x = np.expand_dims(x, axis=0) x = preprocess_input(x) preds = model.predict(x) print('Predicted:') for p in decode_predictions(preds, top=5)[0]: print("Score {}, Label {}".format(p[2],p[1])) ֶशࡁΈϞσϧΛ࢖ͬͨॲཧͷઆ໌ɿͦͷ ը૾ͷલॲཧ

Slide 96

Slide 96 text

from keras.preprocessing import image from keras.applications.inception_v3 \ import preprocess_input, decode_predictions, InceptionV3 import numpy as np model = InceptionV3(weights='imagenet') img_path=' banana.jpg' img = image.load_img(img_path, target_size=(299,299)) x = image.img_to_array(img) x = np.expand_dims(x, axis=0) x = preprocess_input(x) preds = model.predict(x) print('Predicted:') for p in decode_predictions(preds, top=5)[0]: print("Score {}, Label {}".format(p[2],p[1])) ֶशࡁΈϞσϧΛ࢖ͬͨॲཧͷઆ໌ɿͦͷ ਪ࿦

Slide 97

Slide 97 text

from keras.preprocessing import image from keras.applications.inception_v3 \ import preprocess_input, decode_predictions, InceptionV3 import numpy as np model = InceptionV3(weights='imagenet') img_path=' banana.jpg' img = image.load_img(img_path, target_size=(299,299)) x = image.img_to_array(img) x = np.expand_dims(x, axis=0) x = preprocess_input(x) preds = model.predict(x) print('Predicted:') for p in decode_predictions(preds, top=5)[0]: print("Score {}, Label {}".format(p[2],p[1])) ֶशࡁΈϞσϧΛ࢖ͬͨॲཧͷઆ໌ɿͦͷ ਪ࿦݁Ռͷग़ྗ

Slide 98

Slide 98 text

όφφͷ݁Ռ UIPOOZͱ͍͏σϑΥϧτͰ ೖ͍ͬͯΔQZUIPOͷ*%&Λ ࠓճ࢖͍·͢ɻ LFSBTΛ࢖ͬͨը૾ೝࣝΛ ࢼ͢ͷʹTBNQMFQZΛ ։͖·͢ɻ DEd UIPOOZTBNQMFQZ ͱλʔϛφϧͰଧ͍ͬͯͩ͘͞ɻ είΞ όφφͱೝࣝ

Slide 99

Slide 99 text

from keras.preprocessing import image from keras.applications.inception_v3 \ import preprocess_input, decode_predictions, InceptionV3 import numpy as np model = InceptionV3(weights='imagenet') img_path=' apple.jpg' img = image.load_img(img_path, target_size=(299,299)) x = image.img_to_array(img) x = np.expand_dims(x, axis=0) x = preprocess_input(x) preds = model.predict(x) ֶशࡁΈϞσϧΛ࢖ͬͯը૾ೝࣝ͢Δ޻ఔͱ͸ BQQMFKQH ɹˣ ɹߦ໨ͷ CBOBOBKQHΛ ɹBQQMFKQHʹ มߋ͢Δ ɹɹˣ ಉ͡Α͏ʹ࣮ߦ ͯ݁͠ՌΛΈΔ

Slide 100

Slide 100 text

ΓΜ͝ը૾Ͱ࣮ߦ ̍ʣϝχϡʔͷ36/ϘλϯΛԡ͢ ̎ʣ4IFMMͷʮ3VONPEFMQZʯͱදࣔ͞ΕͯϓϩάϥϜ͕࣮ߦ͞Ε·͢ɻ ಉ͡Α͏ʹ̎෼̏̌ඵఔͰ݁Ռ͕ग़·͢ɻͲͷΑ͏ͳ݁ՌʹͳΔͰ͠ΐ͏͔ʁʁ

Slide 101

Slide 101 text

݁Ռ͕ग़Δ·Ͱ̑෼΄Ͳٳܜ

Slide 102

Slide 102 text

ΓΜ͝ͷ݁Ռ είΞ (SBOOZ@4NJUIͱೝࣝ ͋Εʁ BQQMF͡Όͳ͍ͷʁʁ (SBOOZ@4NJUIͱ͸ΓΜ͝ͷ඼छͰ͢ɻֶशͤ͞Δ࣌ͷϥϕϧ͕݁ՌʹͰ·͢ɻ

Slide 103

Slide 103 text

σʔληοτͷ඼࣭ͱ༧ଌ݁Ռ (SBOOZ4NJUI ͜Ε͸౰વ BQQMFͱ ༧ଌ͞ΕΔ ΑͶ ɹσʔληοτͰॻ͔Ε͍ͯΔϥϕϧ ˣ ಛ௃͕Ұ൪ࣅ͍ͯΔͱ൑அ͞ΕΔ ˣ ༧ଌ݁Ռʮ(SBOOZ4NJUIʯ σʔληοτͷ ඼࣭͸ ͱͯ΋ॏཁ

Slide 104

Slide 104 text

࣮ࡍʹΧϝϥͰࡱͬͨ΋ͷΛೝࣝͯ͠Έ·͠ΐ͏ Ұ୴Πϯετʔϧ͢Δͱϝχϡʔ͔Β΋ىಈͰ͖·͢ɻ ϝχϡʔͷɹɹΛΫϦοΫͯ͠ α΢ϯυͱϏσΦʼHVWDWJFX ΛΫϦοΫ͠·͢ɻ

Slide 105

Slide 105 text

64#Χϝϥͷಈ࡞ςετʢࡱͬͨࣸਅΛ֬ೝ͠·͠ΐ͏ʣ อଘ͞ΕͨࣸਅΛΫϦοΫͯ͠ ੩ࢭըΛ֬ೝ͠·͠ΐ͏ɻ ʢNZ@QIPUP൪߸KQHͰ ɹอଘ͞Ε·͢ɻʣ ϑΥϧμϚʔΫΛΫϦοΫͯ͠ QJϑΥϧμΛ։͖·͠ΐ͏ ΩʔϘʔυͷ*Λԡ͢ͱ Ωϟϓνϟʔ͕ࡱΕ·͢

Slide 106

Slide 106 text

ࠓΧϝϥͰࡱͬͨը૾Ͱࢼͯ͠ΈΑ͏ɿ ࡱͬͨը૾ ɹˣ ɹߦ໨ͷ BQQMFKQHΛ ɹը૾໊ʹ มߋ͢Δ ɹɹˣ ಉ͡Α͏ʹ࣮ߦ ͯ݁͠ՌΛΈΔ ‘ࡱͬͨը૾໊’

Slide 107

Slide 107 text

ΧϝϥͰࡱͬͨը૾Ͱ࣮ߦ ̍ʣϝχϡʔͷ36/ϘλϯΛԡ͢ ̎ʣ4IFMMͷʮ3VONPEFMQZʯͱදࣔ͞ΕͯϓϩάϥϜ͕࣮ߦ͞Ε·͢ɻ ಉ͡Α͏ʹ̑෼ඵఔ౓Ͱ݁Ռ͕ग़·͢ɻͲͷΑ͏ͳ݁ՌʹͳΔͰ͠ΐ͏͔ʁʁ

Slide 108

Slide 108 text

3BTQCFSSZ1JͳΒ͜Μͳ͜ͱग़དྷ·͢ʂ ΈΜͳͷϥζύΠίϯςετͰ༏ྑ৆ɹ࡞඼ྫʣ

Slide 109

Slide 109 text

Microsoft Azure Λ࢖༻ͯ͠ সإ൑ఆ Emotion API ௒খܕΧϝϥ RaspberryPi ΧϝϥͰ ࡱӨͨ͠ը૾ 3BTQCFSSZ1JͱػցֶशΛ࢖ͬͨྫɹɹɹɹ 3BTQCFSSZ1JΛ࢖ͬͨ࡞඼ΛԠืͯ͠ΈΑ͏ʂɹɹɹ

Slide 110

Slide 110 text

ΧϝϥͰࡱͬͨը૾ͷ݁Ռ Έͳ͞Μ͸ԿΛ ࡱӨ͠·͔ͨ͠ʁ ݁ՌΛͥͻڭ͑ͯ ͍ͩ͘͞

Slide 111

Slide 111 text

ωοτͰݕࡧͨ͠ը૾Ͱࢼͯ͠ΈΑ͏ Έͳ͞Μ͸ԿΛ ࢼ͠·͔ͨ͠ʁ ݁ՌΛͥͻڭ͑ͯ ͍ͩ͘͞ $ISPNJVNͱ͍͏ϒϥ΢β্ཱ͕͕ͪΔͷͰωοτݕࡧͯ͠Έ͍ͯͩ͘͞ɻ

Slide 112

Slide 112 text

ࠓ೔Έͳ͞Μ͸ֶशϞσϧͷ࡞੒ͱ ࡞੒͞ΕֶͨशϞσϧͰࣗ෼͕ࡱӨͨ͠ ը૾Λ࢖͍ը૾ೝࣝͤ͞Δ͜ͱ͕Ͱ͖·ͨ͠ɻ ͜ΕͰແࣄ3BTQCFSSZ1JͰσΟʔϓϥʔχϯάΛ ͸͡ΊΔ͜ͱ͕Ͱ͖·ͨ͠ʂ ͓ΊͰͱ͏͍͟͝·͢

Slide 113

Slide 113 text

PQFODWΛ࢖ͬͨإೝࣝ

Slide 114

Slide 114 text

5IPOOZͷ-PBEΛΫϦοΫͯ͠ ։͖GBDFQZΛબ୒ͯ͠ ։͖·͢ɻ PQFODWΛ࢖ͬͯإೝࣝΛ͢Δલʹ

Slide 115

Slide 115 text

࣮ߦ ̍ʣϝχϡʔͷ36/ϘλϯΛԡ͢ ̎ʣΧϝϥը૾ͱάϨΠεέʔϧʹมߋͤͨ͞ը૾্ཱ͕͕ͪΓ·͢ ̏ʣࣗ෼ͷإ͕ݟ͑ΔҐஔʹΧϝϥΛஔ͍ͯ֬ೝͰ͖ͨΒ4501Λԡ͢

Slide 116

Slide 116 text

إೝࣝ͢Δલʹը૾ͷ൓సΛ௚ͦ͏ SFU GSBNFDBQSFBE GSBNFDWqJQ GSBNF HSBZDWDWU$PMPS GSBNF DW$0-03@#(3(3": Λ̍ʹ͠·͢ɻ

Slide 117

Slide 117 text

΋͏Ұ౓࣮ߦ ̍ʣϝχϡʔͷ36/ϘλϯΛԡ͢ ̎ʣ൓స͕௚ͬͨը૾͕ݟ͑Δ͸ͣɻ ̏ʣ֬ೝͰ͖ͨΒ4501Λԡ͢

Slide 118

Slide 118 text

5IPOOZͷ-PBEΛΫϦοΫͯ͠ ։͖GBDF̎QZΛબ୒ͯ͠ ։͖·͢ɻ PQFODWΛ࢖ͬͯإೝࣝΛ͢Δલʹ

Slide 119

Slide 119 text

࣮ߦ ̍ʣϝχϡʔͷ36/ϘλϯΛԡ͢ ̎ʣإͷྠֲͱ໨Λೝࣝ͠·͢ɻ

Slide 120

Slide 120 text

Χεέʔυ෼ྨث ࣗ෼ͷإ͕ແࣄೝࣝͰ͖ͨΒFTDΩʔΛԡ͢ɻ΋͘͠͸4501 ͳʹ͔Λೝࣝ͢Δʹ͸ɺͦͷಛ௃Λநग़ͨ͠σʔλ͕ඞཁͰ ࠓճ࢖͍ͬͯΔͷ͕ʮΧεέʔυ෼ྨثʯͰ͢ɻ

Slide 121

Slide 121 text

⭐ਖ਼౴཰͸·ͩ·ׂ͙ͩ̕Β͍ͳͷͰɺਫ਼౓Λ͋͛Δ ࠓޙ͸͜ͷߨ࠲Λडߨͨ͠ܦݧΛੜ͔ͯ͠ ⭐ϦΞϧλΠϜͰ਺ࣈΛೝࣝͰ͖ΔΑ͏ʹ͢Δ ⭐ࣗ෼ͰσʔληοτΛ࡞ͬͯΈΔɹɹɹͳͲͳͲ

Slide 122

Slide 122 text

ߨٛ͸͜ΕͰऴྃͰ͢ɻ ͝੩ௌ͋Γ͕ͱ͏͍͟͝·ͨ͠ɻ