1.2 / July 31, 2018
(3.5/5) (2)
Loading...

Description

This digit recognition application is able to detect andidentifysingle numerical digits (0-9). It served as a greatexperiment thattested the skills I've gained after completingAndrew Ng’s MachineLearning course on Coursera. The 3 layeredneural network wastrained on 5000 training examples (that I havemodified) from theMNIST database of handwritten digits on Octave.Then, I transferredover the test pre-trained weights over toUnity3D, where I havejust implemented forward propagation usingthese weights. In theabout section within the app, I have brieflyoutlined the how therecognizer works while explaining the innerworkings of FeedforwardNeural Networks if you're curious. Uponthorough testing (5minutes), I have come to the conclusion that thedigit recognizer Ihave built is unable to recognize digits at astate of the artperformance. I believe this is due to a lack oftraining data.Considering this project was created from anexperimentalperspective, I have decided to settle for now. I maydecide to addmore data from the MNIST data later in my life. ...Itmay be alittle rough around the edges, but I will always lovemyfirstborn.If you would like to check out the sourcecode:https://github.com/edweenie123/Digit-Recognition-App

App Information Digit Recognizer Neural Network

  • App Name
    Digit Recognizer Neural Network
  • Package Name
    com.edweenoki.DigitRecognizerNeuralNetwork
  • Updated
    July 31, 2018
  • File Size
    23M
  • Requires Android
    Android 4.1 and up
  • Version
    1.2
  • Developer
    edweenoki
  • Installs
    10+
  • Price
    Free
  • Category
    Educational
  • Developer
  • Google Play Link

edweenoki Show More...

Digit Recognizer Neural Network 1.2 APK
edweenoki
This digit recognition application is able to detect andidentifysingle numerical digits (0-9). It served as a greatexperiment thattested the skills I've gained after completingAndrew Ng’s MachineLearning course on Coursera. The 3 layeredneural network wastrained on 5000 training examples (that I havemodified) from theMNIST database of handwritten digits on Octave.Then, I transferredover the test pre-trained weights over toUnity3D, where I havejust implemented forward propagation usingthese weights. In theabout section within the app, I have brieflyoutlined the how therecognizer works while explaining the innerworkings of FeedforwardNeural Networks if you're curious. Uponthorough testing (5minutes), I have come to the conclusion that thedigit recognizer Ihave built is unable to recognize digits at astate of the artperformance. I believe this is due to a lack oftraining data.Considering this project was created from anexperimentalperspective, I have decided to settle for now. I maydecide to addmore data from the MNIST data later in my life. ...Itmay be alittle rough around the edges, but I will always lovemyfirstborn.If you would like to check out the sourcecode:https://github.com/edweenie123/Digit-Recognition-App
Cubey Cubes 1.0.2 APK
edweenoki
★ Explosive particle systems ★ 20+ unique Cubes to unlock ★Manydifferent types of enemies ★ Free rewards every day! ★ Powerupyour hexahedron with an array of upgrades (Explosive +, Damage+,Knockback + and many more.) Edwin Chen, 14 CAN, ON
Loading...