User:Lindseymathew188

= Tensorflow =

Background
Tensorflow is a software library for dataflow programming, originally developed by google. It is a mathlibray and is also used for machine learning applications, like, neutral networks. Tensorflow has a very flexible architecture, which allows it to compute one or more CPUs or GPUs in a desktop, server, or mobile device with a single API. Basically, tensorflow is an open source machine learning framework for everyone.

Introduction
In this science fair project, you will be running a Convolutional Neural Network (like Mixed National Institute of Standards and Technology (MNIST)) with your own dataset of handwritten digits. In this case, faces, were used in place of handwritten digits. MNIST is a simple computer vision dataset. It consists of images of handwritten digits like these:

Each image is a type of face: frown, smile, or indifferent. Then you’re going to “train a model”/create a code to look at the images and predict which type of face appears.

Procedure
1. Install Tensorflow. Go to https://www.tensorflow.org/install/install_mac#installing_with_native_pip and follow the directions for installing Tensorflow with native pip.

2. Divide a piece of paper into 20 boxes, in each box draw variations of one character you want to test. Create three total pieces of paper each with a different character.

3. Scan each of the three pieces of paper and upload them to the computer. Take screenshots of each variation that was drawn and size them to be 28 x 28 pixels. Take screenshot by holding command-shift-4 and highlighting the area to be screenshot. Then, the screenshot will save to the Desktop and in Mac photo viewer select Tools > Adjust Size > select pixels in the drop down menu and change the width and height to 28. (Uncheck the box that says "Scale proportionality" to custom size)

4. Create two new folders, one named "train" and the other "validate". Inside each of these folders create three additional folders, name these three folders each with the names of the three characters to be tested. All these folders are case sensitive, use all lowercase and be sure to spell correctly.

5. Now, in the train folder, place 15 of the sized images into their corresponding folders. Place the remaining 5 sized images into the corresponding folder under validate.

6.