Slide 1

Slide 1 text

Copyright © ABEJA, Inc. All rights reserved. 2019.06.01 (Sat) Gyori Nagafuchi from Japan How to use Deep learning, with Lovely Cat ฅ^•ﻌ•^ฅ

Slide 2

Slide 2 text

Copyright © ABEJA, Inc. All rights reserved. Who am I ? Kyoko Nagafuchi – Gyori Markering @ABEJA, Inc. Favorite AWS Service : Amazon S3 Hobby : Diving, Traveling @Nagafuchik gyori.n KyokoNagafuchi

Slide 3

Slide 3 text

Copyright © ABEJA, Inc. All rights reserved. Everyday with Beer !!

Slide 4

Slide 4 text

Organize JAWS-UG Cloud Girls Group

Slide 5

Slide 5 text

No content

Slide 6

Slide 6 text

CLOUD MARRIGE

Slide 7

Slide 7 text

No content

Slide 8

Slide 8 text

Copyright © ABEJA, Inc. All rights reserved. ABEJA is AI startup We have 2 offices. Singapore WeWork 22 Cross St. Japan Minato-ku, Tokyo

Slide 9

Slide 9 text

WE ARE HIRING !!

Slide 10

Slide 10 text

Copyright © ABEJA, Inc. All rights reserved. Deep Learning ×

Slide 11

Slide 11 text

No content

Slide 12

Slide 12 text

No content

Slide 13

Slide 13 text

No content

Slide 14

Slide 14 text

Harami Misuji

Slide 15

Slide 15 text

Copyright © ABEJA, Inc. All rights reserved. If Misuji or Harami get lost…

Slide 16

Slide 16 text

No content

Slide 17

Slide 17 text

However…. Harami Misuji

Slide 18

Slide 18 text

1. Take photos of cats. 2. Upload to LINE. 3. Deep learning judge Misuji or Harami or other cats. 4. Send back the result on LINE Thank you for finding Misuji, Harami ♡

Slide 19

Slide 19 text

Copyright © ABEJA, Inc. All rights reserved. How to create deep learning bot. 1. Gather cats data 2. Make datasets to annotate 3. Training with Deep Learning Argorithm 4. Develop inference API 5. Develop bot

Slide 20

Slide 20 text

Copyright © ABEJA, Inc. All rights reserved. How to create deep learning bot. 1. Gather cats data 2. Make datasets to annotate 3. Training with Deep Learning Argorithm 4. Develop inference API 5. Develop LINE bot

Slide 21

Slide 21 text

Data needs meaning for deep learning and as you may know it’s really hard to that. ABEJA Platform makes it easy for you to annotate data. Deep Learning need to annotate many data.

Slide 22

Slide 22 text

Cats release me from hard task :Ὂ Harami Misuji Harami Misuji

Slide 23

Slide 23 text

Data needs meaning for deep learning and as you may know it’s really hard to that. ABEJA Platform makes it easy for you to annotate data. It is toooo hard task… but It is the most important thing. Deep Learning need to annotate some data.

Slide 24

Slide 24 text

For examples labels • Cat • Female • Name is Misuji • Calico cat • Cute • Gyori’s friend

Slide 25

Slide 25 text

Copyright © ABEJA, Inc. All rights reserved. How to create deep learning bot. 1. Gather cats data 2. Make datasets to annotate 3. Training with Deep Learning Argorithm 4. Develop inference API 5. Develop LINE bot

Slide 26

Slide 26 text

How to develop inference API ? Just some clicks only. https://abeja-model-zoo.api.abeja.io/deployments/ 1778046646245/services/ser-41b0e70662594fcc

Slide 27

Slide 27 text

Copyright © ABEJA, Inc. All rights reserved. How to create deep learning bot. 1. Gather cats data 2. Make datasets to annotate 3. Training with Deep Learning Argorithm 4. Develop inference API 5. Develop LINE bot

Slide 28

Slide 28 text

Copyright © ABEJA, Inc. All rights reserved. I use AWS Lambda and API Gateway to develop bot. LINE Developer Tool Message API API Gateway Lambda ABEJA Platform Inference API

Slide 29

Slide 29 text

But….

Slide 30

Slide 30 text

JSON parse errors disturb me… :( { "errorMessage": "Unexpected token e in JSON at position 0", "errorType": "SyntaxError", "stackTrace": [ "JSON.parse ()”, "exports.handler (/var/task/index.js:x:x)" ] }

Slide 31

Slide 31 text

Today is Community Meetup

Slide 32

Slide 32 text

Please support me After this presentation. For Misuji and Harami.