20160928-meganeco

「TensorFlowで趣味の画像収集サーバーを作る９月特大号」
TensorFlowによる認識処理の高速化と新データセットでの訓練・評価検証

Transcript

1. C-LIS CO., LTD.
2. 5FOTPS'MPXͰ
   झຯͷը૾ऩूαʔόʔΛ࡞Δ ݄̕ಛେ߸
   
3. C-LIS CO., LTD.  ༗ࢁ
   ܓೋ
   ʢ,FJKJ
   "3*:"."ʣ C-LIS CO., LTD. AndroidΞϓϦ։ൃऀ ػցֶशॳ৺ऀ ΍ͬͯ·ͤΜ 1IPUP
   
   ,PKJ
   .03*(6$)*
   "6/
   $3&"5*7&
   '*3.
4. C-LIS CO., LTD. 
5. C-LIS CO., LTD. 
6. લճ·Ͱͷ͓࿩
7.  IUUQTUFDICPPLGFTUPSH"
8. C-LIS CO., LTD. 
9. C-LIS CO., LTD. IUUQBN[OUPC,3N
10. C-LIS CO., LTD. .FHBOF /PU
11. ޷Έͷ؟ڸ່ͬը૾ΛࣗಈͰऩू͍ͨ͠
12. C-LIS CO., LTD. Ϟσϧ 7((/FUΛࢀߟʹ৞ΈࠐΈ૚ͷ࿈ଓΛ༻͍ͨ$// ʢ$POWPMVUJPOBM
    /FVSBM
    /FUXPSLʣϞσϧ
     DPO
    YY DPO
    YY GD
     QPPM
    Y DPO
    YY DPO
    YY QPPM
    Y GD
     GD
     PVUQVU
    
13. C-LIS CO., LTD. %BUB
    "VHNFOUBUJPO ը૾Λ
    
    
    ʜ
    ౓·ͰɺͦΕͧΕճసͨ͠ը૾Λ࡞੒
    ಡΈࠐΈ࣌ͷॲཧ

    3BOEPN$SPQʢQYதɺQYྖҬΛ͘Γൈ͘ʣ

    3BOEPN'MJQʢԣ࣠ํ޲ʹ൓సʣ

    3BOEPN6Q%PXOʢॎ࣠ํ޲ʹ൓సʣ

    3BOEPN#SJHIUOFTT

    3BOEPN$POUSBTU
    
14. C-LIS CO., LTD. ܇࿅ ֶशΞϧΰϦζϜ
    "EBN
    ֶश཰
    
    ϛχόον
     
15. C-LIS CO., LTD. ݕূ σʔληοτ͔ΒΛςετ༻ͱͯ͠෼཭
    ਖ਼౴཰
    
    ʙ
     
16. Πϥετإݕग़ثʢ'BDF
    %FUFDUPSʣ
17. C-LIS CO., LTD. Πϥετإσʔληοτ ؟ڸ່ͬͱͦ͏Ͱͳ͍΋ͷɻ߹Θͤͯ ຕΛ
 ਖ਼ྫʢʹإʣͱ͢Δʢ͏ͪςετσʔλຕʣ
    ෛྫ͸ɺطଘͷը૾Λࡉ੾Εʹͯ͠ɺإ͕ͳ͍෦෼  ຕΛෛྫͱ͢Δʢ͏ͪςετσʔλຕʣ 
18. 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
19. 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
20. 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
    
21. C-LIS CO., LTD. ܇࿅ ֶशΞϧΰϦζϜ
    "EBN
    ֶश཰
    
    ϛχόον
     
22. C-LIS CO., LTD. ܇࿅ ϛχόονɺ໿ສεςοϓͷ܇࿅Ͱ
 ϩε཰͕
    
    ·Ͱ௿Լ
    ςετσʔλͰͷਖ਼ղ཰ 
23. إݕग़
24. C-LIS CO., LTD.  إ͸Ͳ͜ʹ͋Δʁ
25. C-LIS CO., LTD. 4FMFDUJWF
    4FBSDI 35
    $PSQPSBUJPOͷʮ౜༲͛ϩϘοτʯ
    5FOTPS'MPXษڧձʢ̐ʣ
    ೥݄೔Ͱൃද  IUUQXXXTMJEFTIBSFOFU:VLJ/BLBHBXBUFOTPSqPXSFW
26. C-LIS CO., LTD. 4FMFDUJWF
    4FBSDI "MQBDBࣾʹΑΔ࣮૷
 IUUQCMPHKQBMQBDBBJFOUSZ
    EMJCͷ࣮૷
 IUUQTHJUIVCDPNEBWJTLJOHEMJCCMPCNBTUFS QZUIPO@FYBNQMFTpOE@DBOEJEBUF@PCKFDU@MPDBUJPOTQZ 
27. C-LIS CO., LTD. લճͷ՝୊ ݕग़࣌ؒͷ୹ॖ
    ೝࣝਫ਼౓ͷ޲্ 
28. ݕग़࣌ؒͷ୹ॖ
29. C-LIS CO., LTD. إݕग़ͷखॱ  ީิྖҬͷ੾Γग़͠ SelectiveSearch
30. C-LIS CO., LTD. إݕग़ͷखॱ  إೝࣝ TensorFlow ީิྖҬͷ੾Γग़͠ SelectiveSearch OݸͷީิྖҬ
31. C-LIS CO., LTD. άϥϑͱηογϣϯΛҙࣝ͢Δ  ީิྖҬͷ੾Γग़͠ SelectiveSearch إೝࣝ TensorFlow άϥϑͷ࡞੒ ηογϣϯͷ։࢝ tf.Session() άϥϑͷ࣮ߦ run OݸͷީิྖҬ ը૾σʔλ
    ʴ ը૾σʔλ
32. C-LIS CO., LTD. إೝࣝ TensorFlow άϥϑͱηογϣϯΛҙࣝ͢Δ  άϥϑͷ࡞੒ ηογϣϯͷ։࢝ tf.Session() άϥϑͷ࣮ߦ run άϥϑ
 Ϧηοτ OݸͷྖҬΛ͢΂ͯධՁ OݸͷީิྖҬ ը૾σʔλ
    ʴ
33. C-LIS CO., LTD.  ࠷దͳྖҬΛ୳ࡧʢ3FHSFTTJPOʣ
34. C-LIS CO., LTD. إೝࣝ TensorFlow ຖճάϥϑΛ࡞੒ɾηογϣϯΛ։࢝  άϥϑͷ࡞੒ ηογϣϯͷ։࢝ άϥϑͷ࣮ߦ run άϥϑ
 Ϧηοτ OݸͷྖҬΛ͢΂ͯධՁ OݸͷީิྖҬ ը૾σʔλ
    ʴ
35.  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()

36.  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
37.  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
38. C-LIS CO., LTD. إೝࣝ TensorFlow ީิྖҬͷΈ༩͑Δ  άϥϑͷ࡞੒ ηογϣϯͷ։࢝ άϥϑͷ࣮ߦ run άϥϑ
 Ϧηοτ OݸͷྖҬΛ͢΂ͯධՁ ը૾σʔλ OݸͷީิྖҬ
39. ೝࣝਫ਼౓ͷ޲্
    σʔληοτͷ֦ॆ
40. C-LIS CO., LTD. 5XJUUFS
    "1* ಛఆͷϢʔβʔͷλΠϜϥΠϯʹ౤ߘ͞ΕͨϝσΟΞ ʢը૾ʣΛμ΢ϯϩʔυ
    ϋογϡΛܭࢉͯ͠ॏෳը૾ΛϑΟϧλϦϯά
 ʢαΠζҧ͍ͳͲͷྨࣅը૾͸ফ͖͠Εͳ͍ʣ 
41. C-LIS CO., LTD. Πϥετإσʔλऩूʹ
 ࠷దͳΞΧ΢ϯτ !CPU@FSFDUJPO
    IUUQTUXJUUFSDPNCPU@FSFDUJPO 
42. C-LIS CO., LTD. ࿐ࠎʹੑతͳը૾͸ɺ΄΅ଘࡏ͠ͳ͍ πΠʔτʹඞͣը૾͕ఴ෇͞Ε͍ͯΔ ը૾ʹਓҎ্ͷঁͷࢠͷإؚ͕·Ε͍ͯΔ
43. 5XJUUFS͔Βऔಘͨ͠ϑΝΠϧ਺
     
    ݕग़ͨ͠إྖҬͷ਺ɹɹɹɹɹɹɹɹ  
44. 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
45. C-LIS CO., LTD. 3FHJPO
    $SPQQFS +BWB
    '9
    
    ,PUMJO  IUUQTHJUIVCDPNLFJKJSFHJPO@DSPQQFS
46. %FNP
47. C-LIS CO., LTD. 3FHJPO
    $SPQQFS /FX
    GFBUVSF
    ɾબ୒தͷྖҬΛϑΥʔΧε
    ɾ6OEP
    ɾTFUUJOHTKTPOʹΑΔઃఆ
 ɹϥϕϧ͝ͱͷ࿮ઢ৭
 ɹฤूɺ࡟আͷՄ൱  IUUQTHJUIVCDPNLFJKJSFHJPO@DSPQQFS
48. C-LIS CO., LTD.  ݕग़ͨ͠إྖҬͷ਺ʢॏෳΛ࡟আʣɹɹɹɹɹɹɹɹ  
    إʢਖ਼ྫʣ 
    ޡݕग़ʢෛྫʣ
     
49. 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
50. C-LIS CO., LTD. ৽σʔληοτʹΑΔ܇࿅ ϛχόονɺສεςοϓͷ܇࿅Ͱ
 ϩε཰͕
    
    ·Ͱ௿Լ
    ςετσʔλʹΑΔݕূͰͷਖ਼ղ཰ 
51. 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
52. C-LIS CO., LTD. چσʔληοτͷ܇࿅ ϛχόονɺ໿ສεςοϓͷ܇࿅Ͱ
 ϩε཰͕
    
    ·Ͱ௿Լ
    ςετσʔλͰͷݕূͰͷਖ਼ղ཰ 
53. C-LIS CO., LTD.  ৽Ϟσϧ͸ੑೳ͕ѱ͍ʁ چςετσʔλ چϞσϧ 
     ৽ςετσʔλ ৽Ϟσϧ 
    
54. C-LIS CO., LTD.  ৽چͷςετσʔλΛަ׵ چϞσϧ ৽ςετσʔλ 
     چςετσʔλ ৽Ϟσϧ 
    
55. C-LIS CO., LTD.  ৽چͷςετσʔλΛݕূ چϞσϧ ৽ςετσʔλ 
     چςετσʔλ ৽Ϟσϧ 
     چςετσʔλ چϞσϧ 
     ৽ςετσʔλ ৽Ϟσϧ 
    
56. C-LIS CO., LTD. إೝࣝʹࣦഊͨ͠σʔλΛݕূ
 ʢچςετσʔλʣ  
    OPU@GBDF  GBDF
57. C-LIS CO., LTD. إೝࣝʹࣦഊͨ͠σʔλΛݕূ
 ʢ৽ςετσʔλʣ  
    OPU@GBDF  GBDF
58. ࣍ճ༧ࠂ
59. C-LIS CO., LTD. σʔλͷΫϨϯδϯάʹΑΔೝࣝਫ਼౓ͷมԽΛݟΔ
    ݕग़଎౓Λ͞Βʹվળ͢Δ ࣍ճ༧ࠂ  IUUQTXXXTBLVSBBEKQLPVLBSZPLV
60. 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͸ɺ˜ɺšɺäΛׂѪͯ͠ ͍·͢ɻ