Giter Club home page Giter Club logo

skku-aip-hw2's Introduction

MNIST2Sequence with Tensorflow

0. MNIST Dataset Layout

Train data: 60000 samples

  • Input shape: (28, 28, 1) -> 28x28 Image
  • Label shape: (5) -> Text label

Test data: 10000 samples

  • Input shape: (28, 28, 1) -> 28x28 Image
  • Label shape: (5) -> Text label

1. Model Structure

  • Encoder

    Input

    Layer Output Shape Connected to
    input_1 (?, 28, 28, 1)

    CNN Block 1

    Layer Output Shape Connected to
    conv2d_1 (?, 28, 28, 32) input_1
    activation_1 (?, 28, 28, 32) conv2d_1
    max_pool_1 (?, 14, 14, 32) activation_1

    CNN Block 2

    Layer Output Shape Connected to
    conv2d_2 (?, 14, 14, 64) max_pool_1
    activation_2 (?, 14, 14, 64) conv2d_2
    max_pool_2 (?, 7, 7, 64) activation_2

    Encoder State Output

    Layer Output Shape Connected to
    flatten_1 (?, 128) max_pool_2
    dense_1 (?, 128) flatten_1
    activation_3 (?, 128) dense_1

  • Decoder

    RNN Block

    Layer Output Shape Connected to
    basic_rnn_1 (?, 5)

    Output

    Layer Output Shape Connected to
    dense_2 (?, 5) basic_rnn_1

2. Model Train Result

Train Plot

  • Average train cost: 0.00002, (at 10 epoch)

Sample Prediction with Test Data

Index Label Output Sequence Index Label Output Sequence
1 2 twoPP 26 7 seven
2 1 onePP 27 4 fourP
3 0 zeroP 28 0 zeroP
4 4 fourP 29 1 onePP
5 1 onePP 30 3 three
6 4 fourP 31 1 onePP
7 9 nineP 32 3 three
8 5 fiveP 33 4 fourP
9 9 nineP 34 7 seven
10 0 zeroP 35 2 twoPP
11 6 sixPP 36 7 seven
12 9 nineP 37 1 onePP
13 0 zeroP 38 2 twoPP
14 1 onePP 39 1 onePP
15 5 fiveP 40 1 onePP
16 9 nineP 41 7 seven
17 7 seven 42 4 fourP
18 3 three 43 2 twoPP
19 4 fourP 44 3 three
20 9 nineP 45 5 fiveP
21 6 sixPP 46 1 onePP
22 6 sixPP 47 2 twoPP
23 5 fiveP 48 4 fourP
24 4 fourP 49 4 fourP
25 0 zeroP 50 6 seven

skku-aip-hw2's People

Contributors

yuneg11 avatar

Watchers

avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.