Skip to content

Latest commit

 

History

History
12 lines (12 loc) · 1.35 KB

README.md

File metadata and controls

12 lines (12 loc) · 1.35 KB
Raw

Handwritten digit classifier

To run code from a Unix-based terminal:

  1. Be sure you have Python 2.7 and pip installed on your local machine.
  2. Open your terminal and cd into your home (~) directory.
  3. Clone the repository by typing git clone https://github.com/nicklip/handwritten_digit_classifer.git. Yes, classifier is misspelled as "classifer", I'll fix this at some point.
  4. cd into the repository. Make sure the directory is named handwritten_digit_classifer, otherwise the app will break! Then use requirements.txt to install all necessary Python libraries by typing pip install -r requirements.txt.
  5. Build the machine learning model by typing python image_classifier.py. This takes 9 minutes to run on my 8-core Macbook Pro. This will use all of your cores and so your computer will be slow while it is running.
  6. Start the application by typing export FLASK_APP=ML_app.py hit enter and then type flask run.
  7. Now you can use the app. Open a browser and put http://127.0.0.1:5000/ in the address bar. The main page of the app should now appear. Here's how to use the app:
  • Click 'Choose File', choose which PNG file you'd like to upload. Click upload.
  • You should see 'Upload Successful!', now click on 'See Digit Prediction'
  • A JSON blob consisting of {'prediction' : digit_the_model_predicted} should now be shown.