zishansami102 / cnn-from-scratch Goto Github PK

Description:

Add a function similar to maxpool() in convnet.py for the pooling layer but based on Average Pool method(MNIST folder)

Description:

Currently Xavier Initialization is being used in convnet.py. Write a different function with any other initialization method.

Description:

Add front-end files(html & css) with a very basic structure of the digit recognition app(explained in README) from where the development can start.

Description:

An API which receives image as input and gives digit prediction and its probability for that image as output(from functions available in 'convnet.py' file).
Use 'app.py' file to start from.
[read the README for more details]

I have a question, how to handle non-number prediction?

I tried your Demo and when I wrote C, it was predicted as 5 and when I wrote A, it gave 2.

How to extend the NN so that it can make an output "Unknown" / "Not in the list"?

I am using Python 3.7 on MacOS.
I trained the CNN with the MNIST db and tried running app.py
While it starts and I can access it from 127.0.0.1:5000, as soon as I draw a digit and press Predict I get the following console output.

• Serving Flask app "app" (lazy loading)
• Environment: production
  WARNING: Do not use the development server in a production environment.
  Use a production WSGI server instead.
• Debug mode: on
• Running on http://127.0.0.1:5000/ (Press CTRL+C to quit)
• Restarting with stat
• Debugger is active!
• Debugger PIN: 162-250-119
  127.0.0.1 - - [25/Aug/2018 00:29:18] "GET / HTTP/1.1" 200 -
  127.0.0.1 - - [25/Aug/2018 00:29:18] "GET /static/style/drawingboard.min.css HTTP/1.1" 200 -
  127.0.0.1 - - [25/Aug/2018 00:29:18] "GET /static/style/application2.min.css HTTP/1.1" 200 -
  127.0.0.1 - - [25/Aug/2018 00:29:18] "GET /static/style/drawingBoard.min.js HTTP/1.1" 200 -
  127.0.0.1 - - [25/Aug/2018 00:29:18] "GET /static/style/hermiteResize.js HTTP/1.1" 200 -
  127.0.0.1 - - [25/Aug/2018 00:29:23] "POST /digit_process HTTP/1.1" 500 -
  Traceback (most recent call last):
  File "/usr/local/lib/python3.7/site-packages/flask/app.py", line 2309, in call
  return self.wsgi_app(environ, start_response)
  File "/usr/local/lib/python3.7/site-packages/flask/app.py", line 2295, in wsgi_app
  response = self.handle_exception(e)
  File "/usr/local/lib/python3.7/site-packages/flask/app.py", line 1741, in handle_exception
  reraise(exc_type, exc_value, tb)
  File "/usr/local/lib/python3.7/site-packages/flask/_compat.py", line 35, in reraise
  raise value
  File "/usr/local/lib/python3.7/site-packages/flask/app.py", line 2292, in wsgi_app
  response = self.full_dispatch_request()
  File "/usr/local/lib/python3.7/site-packages/flask/app.py", line 1815, in full_dispatch_request
  rv = self.handle_user_exception(e)
  File "/usr/local/lib/python3.7/site-packages/flask/app.py", line 1718, in handle_user_exception
  reraise(exc_type, exc_value, tb)
  File "/usr/local/lib/python3.7/site-packages/flask/_compat.py", line 35, in reraise
  raise value
  File "/usr/local/lib/python3.7/site-packages/flask/app.py", line 1813, in full_dispatch_request
  rv = self.dispatch_request()
  File "/usr/local/lib/python3.7/site-packages/flask/app.py", line 1799, in dispatch_request
  return self.view_functionsrule.endpoint
  File "/Users/dennisfurrer/Desktop/CNN-from-Scratch/app.py", line 23, in digit_process
  out = pickle.load(pickle_in)
  UnicodeDecodeError: 'ascii' codec can't decode byte 0xca in position 0: ordinal not in range(128)

How to enable training on GPU's ?
Any pointers in that direction?

pointer
This is a little vague pointer as it will depend on where you are trying to put the batch_norm layer.
Everything depends on what is working and what is not and what is the conventional way to it.