Classifier not so from scratch with GoCV and TensorFlow Hub

Slide 1

Slide 1 text

Classiﬁer not so from scratch with GoCV and TensorFlow Hub Florianópolis - 2019

Slide 2

Slide 2 text

Sheimy Rahman

Slide 3

Slide 3 text

https://www.zenvia.com

Slide 4

Slide 4 text

O que é Computer Vision (CV) ?

Slide 5

Slide 5 text

O que é Computer Vision (CV) ?

Slide 6

Slide 6 text

O que você pode fazer com CV?

Slide 7

Slide 7 text

O que você pode fazer com CV? RECONHECER PESSOAS/FACES DETECTAR MOVIMENTOS TELEPRESENÇA VEÍCULOS AUTÔNOMOS MELHORAMENTO HUMANO

Slide 8

Slide 8 text

O que você pode fazer com CV?

Slide 9

Slide 9 text

E o que Go, a linguagem mais amada, tem a ver com isso?

Slide 10

Slide 10 text

THE GOPHER CAN SEE YOU NOW. gocv.io opencv.org golang.org

Slide 11

Slide 11 text

Por que GO? ➢ Concorrência ➢ Portabilidade ➢ Desempenho

Slide 12

Slide 12 text

Como funciona GoCV package? GO CGO C C++

Slide 13

Slide 13 text

OpenCV GoCV Go Application GoCv Go functions GoCv C wrapper functions

Slide 14

Slide 14 text

GoCV package: ➢ Linux ➢ Mac OS ➢ Windows

Slide 15

Slide 15 text

www.tensorﬂow.org/hub

Slide 16

Slide 16 text

TensorFlow Hub is a library for reusable machine learning modules.

Slide 17

Slide 17 text

A module is a self-contained piece of a TensorFlow graph, along with its weights and assets, that can be reused across different tasks in a process known as transfer learning. Transfer learning can: ● Train a model with a smaller dataset, ● Improve generalization, and ● Speed up training. Ou seja... Modules?

Slide 18

Slide 18 text

CENTENAS DE MODELOS MATEMÁTICOS!!!

Slide 19

Slide 19 text

Laplacians Scharr - Sobel-Feldman operator Absolute Diff/Threshould

Slide 20

Slide 20 text

GoCV + TensorFlowHub

Slide 21

Slide 21 text

Into...

Slide 22

Slide 22 text

gocv.io gobot.io tinygo.org PROJETOS OPEN SOURCE

Slide 23

Slide 23 text

Thanks, @deadprogram for the GoCV stickers! GoCV, I’m your Father!

Slide 24

Slide 24 text

@golangbridge gobridge.org go-store.io

Slide 25

Slide 25 text

OBRIGADA! :) @Sheimy_Rahman [email protected] github.com/Wdrwoman