jmandarino / meter-reader

Graph data from a series of images. Using OCR, KNN/SVM and MINIST

Seven segment display reader (Meter Reader)

Purpose:

To read the data from Coned Electric meters (7 segment displays) and graph the information

Workflow:

1. place images inside in/imgs
2. run setup.py python setup.py this creates images.json
3. process image data from images.json by running python digit_ocr.py
4. run data through MNIST and SVM/KNN
5. Graph data and find trends

TOOLS:

1. opencv
2. sklearn
3. MNIST Dataset
4. Matplotlib

MAJOR TODO's :

1. Connect user flows (processing img -> figure out image -> graph)
2. Error Processing for (OCR)
3. Error Processing for data (if ocr failed to read data correctly what should we do
4. centralize paths instead of declaring them in each file

Requirements and how to install

TODO: later

Major issues:

image processing is strong enough given every picture is varied.

meter-reader's Issues

contour issue

Hi.
Thanks for the great project but I ran into the bug. I started the project as you described in the readme but there was an error:
digits_ocr.py:

displayCnt = None
....
lines 57-58:
if len(approx) == 4:
displayCnt = approx
displayCnt stays None!
and therefore
warped = four_point_transform(gray, displayCnt.reshape(4, 2))
doesn't work.

Could you please to fix that issue.

Thanks for your cool work again,
Andrii.