20160928-meganeco - Speaker Deck

20160928-meganeco

by ARIYAMA Keiji

Slide 1

Slide 1 text

C-LIS CO., LTD.

Slide 2

Slide 2 text

5FOTPS'MPXͰ झຯͷը૾ऩूαʔόʔΛ࡞Δ ݄̕ಛେ߸

Slide 3

Slide 3 text

C-LIS CO., LTD. ༗ࢁܓೋʢ,FJKJ"3*:"."ʣ C-LIS CO., LTD. AndroidΞϓϦ։ൃऀ ػցֶशॳ৺ऀ ΍ͬͯ·ͤΜ 1IPUP,PKJ.03*(6$)* "6/$3&"5*7&'*3.

Slide 4

Slide 4 text

C-LIS CO., LTD.

Slide 5

Slide 5 text

C-LIS CO., LTD.

Slide 6

Slide 6 text

લճ·Ͱͷ͓࿩

Slide 7

Slide 7 text

IUUQTUFDICPPLGFTUPSH"

Slide 8

Slide 8 text

C-LIS CO., LTD.

Slide 9

Slide 9 text

C-LIS CO., LTD. IUUQBN[OUPC,3N

Slide 10

Slide 10 text

C-LIS CO., LTD. .FHBOF /PU

Slide 11

Slide 11 text

޷Έͷ؟ڸ່ͬը૾ΛࣗಈͰऩू͍ͨ͠

Slide 12

Slide 12 text

C-LIS CO., LTD. Ϟσϧ 7((/FUΛࢀߟʹ৞ΈࠐΈ૚ͷ࿈ଓΛ༻͍ͨ$// ʢ$POWPMVUJPOBM/FVSBM/FUXPSLʣϞσϧ DPO YY DPO YY GD QPPM Y DPO YY DPO YY QPPM Y GD GD PVUQVU

Slide 13

Slide 13 text

C-LIS CO., LTD. %BUB"VHNFOUBUJPO ը૾Λ ʜ౓·ͰɺͦΕͧΕճసͨ͠ը૾Λ࡞੒ ಡΈࠐΈ࣌ͷॲཧ  3BOEPN$SPQʢQYதɺQYྖҬΛ͘Γൈ͘ʣ  3BOEPN'MJQʢԣ࣠ํ޲ʹ൓సʣ  3BOEPN6Q%PXOʢॎ࣠ํ޲ʹ൓సʣ  3BOEPN#SJHIUOFTT  3BOEPN$POUSBTU

Slide 14

Slide 14 text

C-LIS CO., LTD. ܇࿅ ֶशΞϧΰϦζϜ"EBN ֶश཰ ϛχόον

Slide 15

Slide 15 text

C-LIS CO., LTD. ݕূ σʔληοτ͔ΒΛςετ༻ͱͯ͠෼཭ ਖ਼౴཰ʙ

Slide 16

Slide 16 text

Πϥετإݕग़ثʢ'BDF%FUFDUPSʣ

Slide 17

Slide 17 text

C-LIS CO., LTD. Πϥετإσʔληοτ ؟ڸ່ͬͱͦ͏Ͱͳ͍΋ͷɻ߹Θͤͯ ຕΛ  ਖ਼ྫʢʹإʣͱ͢Δʢ͏ͪςετσʔλຕʣ ෛྫ͸ɺطଘͷը૾Λࡉ੾Εʹͯ͠ɺإ͕ͳ͍෦෼ ຕΛෛྫͱ͢Δʢ͏ͪςετσʔλຕʣ

Slide 18

Slide 18 text

C-LIS CO., LTD. Πϥετσʔληοτ ਖ਼ྫ ෛྫ ߹ܭ ܇࿅σʔλ 1,600 3,200 4,800 ςετσʔλ 400 800 1,200 ߹ܭ 2,000 4,005 6,000

Slide 19

Slide 19 text

C-LIS CO., LTD. Ϟσϧ $*'"3νϡʔτϦΞϧͷϞσϧΛࢀߟʹ  υϩοϓΞ΢τʢˋʣΛ௥Ճͨ͠$//ʢ$POWPMVUJPOBM/FVSBM /FUXPSLʣ DPO YY GD QPPM Y DPO YY GD PVUQVU MSO MSO QPPM Y

Slide 20

Slide 20 text

C-LIS CO., LTD. %BUB"VHNFOUBUJPO ը૾͸Yʹॖখ ը૾Λ ʜ౓ɺͦΕͧΕճసͨ͠ը૾Λ࡞੒ ಡΈࠐΈ࣌ͷॲཧ  3BOEPN$SPQʢQYதɺQYྖҬͰ͘Γൈ͘ʣ  3BOEPN'MJQʢԣ࣠ํ޲ʹ൓సʣ  3BOEPN6Q%PXOʢॎ࣠ํ޲ʹ൓సʣ  3BOEPN#SJHIUOFTT  3BOEPN$POUSBTU

Slide 21

Slide 21 text

C-LIS CO., LTD. ܇࿅ ֶशΞϧΰϦζϜ"EBN ֶश཰ ϛχόον

Slide 22

Slide 22 text

C-LIS CO., LTD. ܇࿅ ϛχόονɺ໿ສεςοϓͷ܇࿅Ͱ  ϩε཰͕·Ͱ௿Լ ςετσʔλͰͷਖ਼ղ཰

Slide 23

Slide 23 text

إݕग़

Slide 24

Slide 24 text

C-LIS CO., LTD. إ͸Ͳ͜ʹ͋Δʁ

Slide 25

Slide 25 text

C-LIS CO., LTD. 4FMFDUJWF4FBSDI 35$PSQPSBUJPOͷʮ౜༲͛ϩϘοτʯ 5FOTPS'MPXษڧձʢ̐ʣ೥݄೔Ͱൃද IUUQXXXTMJEFTIBSFOFU:VLJ/BLBHBXBUFOTPSqPXSFW

Slide 26

Slide 26 text

C-LIS CO., LTD. 4FMFDUJWF4FBSDI "MQBDBࣾʹΑΔ࣮૷  IUUQCMPHKQBMQBDBBJFOUSZ EMJCͷ࣮૷  IUUQTHJUIVCDPNEBWJTLJOHEMJCCMPCNBTUFS QZUIPO@FYBNQMFTpOE@DBOEJEBUF@PCKFDU@MPDBUJPOTQZ

Slide 27

Slide 27 text

C-LIS CO., LTD. લճͷ՝୊ ݕग़࣌ؒͷ୹ॖ ೝࣝਫ਼౓ͷ޲্

Slide 28

Slide 28 text

ݕग़࣌ؒͷ୹ॖ

Slide 29

Slide 29 text

C-LIS CO., LTD. إݕग़ͷखॱ ީิྖҬͷ੾Γग़͠ SelectiveSearch

Slide 30

Slide 30 text

C-LIS CO., LTD. إݕग़ͷखॱ إೝࣝ TensorFlow ީิྖҬͷ੾Γग़͠ SelectiveSearch OݸͷީิྖҬ

Slide 31

Slide 31 text

C-LIS CO., LTD. άϥϑͱηογϣϯΛҙࣝ͢Δ ީิྖҬͷ੾Γग़͠ SelectiveSearch إೝࣝ TensorFlow άϥϑͷ࡞੒ ηογϣϯͷ։࢝ tf.Session() άϥϑͷ࣮ߦ run OݸͷީิྖҬ ը૾σʔλʴ ը૾σʔλ

Slide 32

Slide 32 text

C-LIS CO., LTD. إೝࣝ TensorFlow άϥϑͱηογϣϯΛҙࣝ͢Δ άϥϑͷ࡞੒ ηογϣϯͷ։࢝ tf.Session() άϥϑͷ࣮ߦ run άϥϑ  Ϧηοτ OݸͷྖҬΛ͢΂ͯධՁ OݸͷީิྖҬ ը૾σʔλʴ

Slide 33

Slide 33 text

C-LIS CO., LTD. ࠷దͳྖҬΛ୳ࡧʢ3FHSFTTJPOʣ

Slide 34

Slide 34 text

C-LIS CO., LTD. إೝࣝ TensorFlow ຖճάϥϑΛ࡞੒ɾηογϣϯΛ։࢝ άϥϑͷ࡞੒ ηογϣϯͷ։࢝ άϥϑͷ࣮ߦ run άϥϑ  Ϧηοτ OݸͷྖҬΛ͢΂ͯධՁ OݸͷީิྖҬ ը૾σʔλʴ

Slide 35

Slide 35 text

class FaceDetector(object):  image_path = None  original_image = None  graph = None  sess = None  queue = None  top_k_indices = None  top_k_values = None  region_batch = None  coord = None    configuration = None    def __init__(self, image_path, batch_size, train_dir):  self.image_path = image_path  self.original_image = Image.open(image_path)  self.original_image = self.original_image.convert('RGB')    checkpoint = tf.train.get_checkpoint_state(train_dir)  if not (checkpoint and checkpoint.model_checkpoint_path):  print('νΣοΫϙΠϯτϑΝΠϧ͕ݟ͔ͭΓ·ͤΜ')  return    self.graph, self.queue, self.top_k_indices, self.top_k_values, self.region_batch = \  self._init_graph(self.original_image, batch_size)    with self.graph.as_default() as g:  self.sess = tf.Session()  saver = tf.train.Saver()  saver.restore(self.sess, checkpoint.model_checkpoint_path)    self.coord = tf.train.Coordinator() 

Slide 36

Slide 36 text

def _init_graph(self, image, batch_size):  reshaped_image = np.array(image.getdata()).reshape(image.height, image.width, 3).astype(  np.float32)    graph = tf.Graph()    with graph.as_default() as g:  queue = tf.FIFOQueue(3000, tf.int32, shapes=[4])    region = queue.dequeue()  whitten_image = self._load_image(reshaped_image, region)    image_batch, region_batch = tf.train.batch(  [whitten_image, region],  batch_size=batch_size,  capacity=10000)    logits = tf.nn.softmax(model.inference(image_batch, tf.constant(1.0), batch_size))  top_k_values, top_k_indices = tf.nn.top_k(logits, 2, sorted=True)    return graph, queue, top_k_indices, top_k_values, region_batch

Slide 37

Slide 37 text

def _eval(self, region_list, batch_size):  result = []    with graph.as_default() as g:  threads = tf.train.start_queue_runners(sess=sess, coord=coord)    step = 0  try:  # όοναΠζʹ߹Θͤͯ਺Λௐ੔  while len(region_list) < batch_size or len(region_list) % batch_size != 0:  add = region_list[0:(batch_size - (len(region_list) % batch_size))]  region_list = region_list + add    region_list = np.array(region_list)    enqueue = queue.enqueue_many(region_list)  sess.run(enqueue)    num_iter = int(math.ceil(len(region_list) / batch_size))    while step < num_iter and not coord.should_stop():    # ҎԼུ

Slide 38

Slide 38 text

C-LIS CO., LTD. إೝࣝ TensorFlow ީิྖҬͷΈ༩͑Δ άϥϑͷ࡞੒ ηογϣϯͷ։࢝ άϥϑͷ࣮ߦ run άϥϑ  Ϧηοτ OݸͷྖҬΛ͢΂ͯධՁ ը૾σʔλ OݸͷީิྖҬ

Slide 39

Slide 39 text

ೝࣝਫ਼౓ͷ޲্ σʔληοτͷ֦ॆ

Slide 40

Slide 40 text

C-LIS CO., LTD. 5XJUUFS"1* ಛఆͷϢʔβʔͷλΠϜϥΠϯʹ౤ߘ͞ΕͨϝσΟΞ ʢը૾ʣΛμ΢ϯϩʔυ ϋογϡΛܭࢉͯ͠ॏෳը૾ΛϑΟϧλϦϯά  ʢαΠζҧ͍ͳͲͷྨࣅը૾͸ফ͖͠Εͳ͍ʣ

Slide 41

Slide 41 text

C-LIS CO., LTD. Πϥετإσʔλऩूʹ  ࠷దͳΞΧ΢ϯτ !CPU@FSFDUJPO IUUQTUXJUUFSDPNCPU@FSFDUJPO

Slide 42

Slide 42 text

C-LIS CO., LTD. ࿐ࠎʹੑతͳը૾͸ɺ΄΅ଘࡏ͠ͳ͍ πΠʔτʹඞͣը૾͕ఴ෇͞Ε͍ͯΔ ը૾ʹਓҎ্ͷঁͷࢠͷإؚ͕·Ε͍ͯΔ

Slide 43

Slide 43 text

5XJUUFS͔Βऔಘͨ͠ϑΝΠϧ਺ ݕग़ͨ͠إྖҬͷ਺ɹɹɹɹɹɹɹɹ

Slide 44

Slide 44 text

C-LIS CO., LTD. { "detected_faces": { "mode": "selective_search", "regions": [ { "label": 1, "rect": { "left": 212.0, "bottom": 654.0, "top": 94.0, "right": 483.0 }, "probability": 0.9994481205940247 } ] }, "created_at": "2016-09-28T03:37:16.223942", "file_name": "kage_maturi-CAARxE8UwAEzAjJ.jpg", "generator": "Megane Co" } ݁ՌΛ+40/Ͱॻ͖ग़͠

Slide 45

Slide 45 text

C-LIS CO., LTD. 3FHJPO$SPQQFS +BWB'9,PUMJO IUUQTHJUIVCDPNLFJKJSFHJPO@DSPQQFS

Slide 46

Slide 46 text

%FNP

Slide 47

Slide 47 text

C-LIS CO., LTD. 3FHJPO$SPQQFS /FXGFBUVSF ɾબ୒தͷྖҬΛϑΥʔΧε ɾ6OEP ɾTFUUJOHTKTPOʹΑΔઃఆ  ɹϥϕϧ͝ͱͷ࿮ઢ৭  ɹฤूɺ࡟আͷՄ൱ IUUQTHJUIVCDPNLFJKJSFHJPO@DSPQQFS

Slide 48

Slide 48 text

C-LIS CO., LTD. ݕग़ͨ͠إྖҬͷ਺ʢॏෳΛ࡟আʣɹɹɹɹɹɹɹɹ إʢਖ਼ྫʣ ޡݕग़ʢෛྫʣ

Slide 49

Slide 49 text

C-LIS CO., LTD. ৽σʔληοτ ਖ਼ྫ ෛྫ ߹ܭ ܇࿅σʔλ 10,453 10,792 21,245 ςετσʔλ 2,619 2,703 5,322 ߹ܭ 13,072 13,495 26,567

Slide 50

Slide 50 text

C-LIS CO., LTD. ৽σʔληοτʹΑΔ܇࿅ ϛχόονɺສεςοϓͷ܇࿅Ͱ  ϩε཰͕·Ͱ௿Լ ςετσʔλʹΑΔݕূͰͷਖ਼ղ཰

Slide 51

Slide 51 text

C-LIS CO., LTD. چσʔληοτ ਖ਼ྫ ෛྫ ߹ܭ ܇࿅σʔλ 1,600 3,200 4,800 ςετσʔλ 400 800 1,200 ߹ܭ 2,000 4,005 6,000

Slide 52

Slide 52 text

C-LIS CO., LTD. چσʔληοτͷ܇࿅ ϛχόονɺ໿ສεςοϓͷ܇࿅Ͱ  ϩε཰͕·Ͱ௿Լ ςετσʔλͰͷݕূͰͷਖ਼ղ཰

Slide 53

Slide 53 text

C-LIS CO., LTD. ৽Ϟσϧ͸ੑೳ͕ѱ͍ʁ چςετσʔλ چϞσϧ ৽ςετσʔλ ৽Ϟσϧ

Slide 54

Slide 54 text

C-LIS CO., LTD. ৽چͷςετσʔλΛަ׵ چϞσϧ ৽ςετσʔλ چςετσʔλ ৽Ϟσϧ

Slide 55

Slide 55 text

C-LIS CO., LTD. ৽چͷςετσʔλΛݕূ چϞσϧ ৽ςετσʔλ چςετσʔλ ৽Ϟσϧ چςετσʔλ چϞσϧ ৽ςετσʔλ ৽Ϟσϧ

Slide 56

Slide 56 text

C-LIS CO., LTD. إೝࣝʹࣦഊͨ͠σʔλΛݕূ  ʢچςετσʔλʣ OPU@GBDF GBDF

Slide 57

Slide 57 text

C-LIS CO., LTD. إೝࣝʹࣦഊͨ͠σʔλΛݕূ  ʢ৽ςετσʔλʣ OPU@GBDF GBDF

Slide 58

Slide 58 text

࣍ճ༧ࠂ

Slide 59

Slide 59 text

C-LIS CO., LTD. σʔλͷΫϨϯδϯάʹΑΔೝࣝਫ਼౓ͷมԽΛݟΔ ݕग़଎౓Λ͞Βʹվળ͢Δ ࣍ճ༧ࠂ IUUQTXXXTBLVSBBEKQLPVLBSZPLV

Slide 60

Slide 60 text

C-LIS CO., LTD. C-LIS CO., LTD. ຊࢿྉ͸ɺ༗ݶձࣾγʔϦεͷஶ࡞෺Ͱ͢ɻܝࡌ͞Ε͍ͯΔΠϥετʹ͸ݸผʹஶ࡞ݖ͕͋Γ·͢ɻ ຊࢿྉͷશ෦ɺ·ͨ͸Ұ෦ʹ͍ͭͯɺஶ࡞ऀ͔ΒจॻʹΑΔڐ୚Λಘͣʹෳ੡͢Δ͜ͱ͸ې͡ΒΕ͍ͯ·͢ɻ ֤੡඼໊ɾϒϥϯυ໊ɺձ໊ࣾͳͲ͸ɺҰൠʹ֤ࣾͷ঎ඪ·ͨ͸ొ࿥঎ඪͰ͢ɻຊࢿྉதͰ͸ɺɺɺäΛׂѪͯ͠ ͍·͢ɻ