Practice-ML-with-Digit-recognize-problem

Solving digits recognize problem

Introdution:

As the name, I think that joining into the problem, analyzing and coding to solve it is the easiest way to improve knowledge of Maching learning - Deep learning and solving problem skills.

In this problem we just recognize the handwritten digit, not detect. I use Colab to train the models.

Resource:

I use two datasets:

• Kaggle: mnist-dataset -- You have to download, put into a drive folder and change some url in train_model file to run.
• Keras: mnist-dataset -- You just run cell to load the dataset.

Explanation of files:

1. In code folder:
   • Train_model1.ipynb: train_file with dataset from Kaggle
   • Train_model2.ipynb: train_file with dataset from Keras.
   • detect.py: to identifies a digit of an image.

      > cd code
      > detect.py --source ../data/test_image1.png
     

2. In data folder:
   • saved_model1.h5 and saved_model1.pkl: weight of model1.
   • saved_model2.h5 and saved_model2.pkl: weight of model2.
   • test_image.png: image to test in applicate_model.py