Implementing program to run own data recognition

Is there a way to run out own data so the model will read our own input data for handwritten documents?

@psantillo this is also possible but you will need to upload images and predict against the uploaded images using the trained images here is an example of uploading images to drive.

import numpy as np
from google.colab import files
from keras.preprocessing import image

im_upload = files.upload()

for fn in im_upload.keys():

  path = fn
  img = image.load_img(path, target_size=(100, 100))
  x = image.img_to_array(img)
  x = np.expand_dims(x, axis=0)

  images = np.vstack([x])
  # Predictions can be done here
  print(x.shape) # (1, 100, 100, 3)

You can customize the image size target according to the training model.

Thank you for that, would it be possible to train against my own images and MNIST data set?

1 Like

Yeah it is possible but you must follow all the steps of preparing the training Data set and I think it can take you time to collect all Images and organize them for the training process.

And it is a process that requires expertise as you need to clearly understand the data quality you need.
But it is also a possible alternative that needs some research on how Data sets are made.
Preparing Your Dataset for Machine Learning: 10 Basic Techniques That Make Your Data Better
I hope this article can help you to learn more.


this helps thank you very much Hirwa!

1 Like