Digits Recognition using Machine Learning Algorithm | Python, sklearn




[ad_1]

Digits recognition using python machine learning algorithm.

*Note: This is just a basic approach to understand Machine Learning with image processing. This does not covers advanced real world problems that can be solved by ML, but gives any starter in ML a high level conceptual view of “How ML actually works?”.

Github code of video here https://github.com/arneec/digits-recognition/tree/v1.0.

Github code for latest library supports and requirements details here https://github.com/arneec/digits-recognition.

** Apologizes for my voice and audio quality. **

Source


[ad_2]

Comment List

  • Amit Garu
    December 14, 2020

    Hello all, the old version of code has been updated with latest library support and all requirement details and steps to run the program. You can find the the latest code and description @ https://github.com/arneec/digits-recognition .

  • Amit Garu
    December 14, 2020

    Can you give me the dataset?

  • Amit Garu
    December 14, 2020

    Hi Amit,

    I am using PIL to load the image, and facing issue. Please help.

    I am loading the image , and reshaping it to 28*28.. where as when converting it to numpy array at-time it convert into 28*28*3 and at-times into 28*28*4… how to standardize it.

    Below is the code.

    `from PIL import Image

    import numpy as np

    size = 28, 28

    img = Image.open("handwritten_image_256x256.png")

    img

    img = img.resize(size, Image.ANTIALIAS)

    display(img.size)

    img

    img_array = np.array(img)

    display(img_array.shape, img_array)`

  • Amit Garu
    December 14, 2020

    Hello,
    if I change the colors , I mean the numbers are written in black on a white background , the prediction is always [1].
    What should I do in this case ?
    Do I have to change my dataset ?
    Thank you

  • Amit Garu
    December 14, 2020

    hello sir..
    bytescale function is not working in recent versions of python.
    is there any alternative for this.?

  • Amit Garu
    December 14, 2020

    senk man, you realy help me, good like

  • Amit Garu
    December 14, 2020

    What i must install to run this program?

  • Amit Garu
    December 14, 2020

    Bhai itna jaan par arha tha bolne mai toh video q banya…voice is very bad..

  • Amit Garu
    December 14, 2020

    This is best among all the videos… I spent a hell lot of time finding a good project… this one rocked.. but look at your voice and the sounds you are making…..like you are having sex….

  • Amit Garu
    December 14, 2020

    thanks for the tutorial sir
    I have a question though,when we are inputting our own images like you did,are we training the model,is it becoming more accurate as we test it more?

  • Amit Garu
    December 14, 2020

    Is there any alternative for scipy.misc.bytescale?

  • Amit Garu
    December 14, 2020

    First of all thanks for the informative video.
    I am getting this error –
    DeprecationWarning: the imp module is deprecated in favour of importlib; see the module's documentation for alternative uses
    import imp

    With regards

  • Amit Garu
    December 14, 2020

    Graicas Brother……mejora el IA del audio 🙂

  • Amit Garu
    December 14, 2020

    Hello, your voice is very low in sound.. could you speak loud

  • Amit Garu
    December 14, 2020

    it is showing the error that,imread is not present in MISC.

  • Amit Garu
    December 14, 2020

    Which is the compiler used here ?

  • Amit Garu
    December 14, 2020

    how to train this model own data

  • Amit Garu
    December 14, 2020

    Traceback (most recent call last):
    File "devanagari.py", line 27, in <module>
    x_test.append(sum(eachPixel)/3.0)
    TypeError: 'numpy.uint8' object is not iterable
    how can slove this error

  • Amit Garu
    December 14, 2020

    Hi there, first thanks for the video it really helps… I'd like to ask you something. When drawing images on Paint it throws a prediction of what you're drawing but when I try to do the same but with an image taken by a webcam it predicts most of the time [1] even though I follow the same process as on Paint(I rescale the image and so on), my question is: Is it possible to recognize handwritten but with digits taken by a camera?

  • Amit Garu
    December 14, 2020

    Again Sir How are you.
    will you send me "opencv books for beginner level" to read

  • Amit Garu
    December 14, 2020

    I want to detect numbers from national identification card.
    how can I do?

  • Amit Garu
    December 14, 2020

    Thanks bro for the informative video.
    I have one doubt. Is it possible to do the same kind of recognition with hand drawn patterns like signatures? What changes would I need to make for this? Please guide me bro.
    Thanks in advance.

  • Amit Garu
    December 14, 2020

    Hey brother… Thanx a lot for that video…
    I want to ask something.

    Here you have loaded the digits.csv file that already exists in datasets folder.

    How can I use my own file?? I am using the MNIST_TRAIN.CSV file to train my data. Can you please tell some resource to proceed with it?

    Also, I want to save the data of the test image to a CSV file and append it as and when required, so that whenever I open a new file, its array data is saved to one file.

    Expecting an early response.

    Thanks in advance.

  • Amit Garu
    December 14, 2020

    Thanks alot..! it is working absolutely correct 🙂

  • Amit Garu
    December 14, 2020

    Thankss a loott…really efficient…

Write a comment