https://github.com/sarfata/meterreader

Use python+opencv to read an old-school gazmeter
https://github.com/sarfata/meterreader

Last synced: 4 days ago
JSON representation

Use python+opencv to read an old-school gazmeter

• Host: GitHub
• URL: https://github.com/sarfata/meterreader
• Owner: sarfata
• Created: 2018-11-16T23:58:25.000Z (about 6 years ago)
• Default Branch: master
• Last Pushed: 2019-12-17T19:22:40.000Z (about 5 years ago)
• Last Synced: 2024-11-08T15:00:30.419Z (about 2 months ago)
• Language: Python
• Size: 48.8 KB
• Stars: 2
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Gazmeter

## Installation

### MacOS

    brew install python3

    brew postinstall python3 (for pip)

    pip3 install -r requirements.txt (will install opencv-python prebuilt binaries)

    python3 / import cv2 / cv2.__version__ => 4.1.1

## Using

Get images from Amazon:

    aws s3 cp s3://metering-vanves/img img --recursive --exclude '*' --include 'image-2019*-0900*jpg'

Experiment:

    ./meterreader.py experiment img201911/*jpg

Label some images:

    python3 meterreader.py label-samples img

For each image shown, click on the image window and type the digit you see. Press enter when done. This will add a txt file next to each image with the value.

Test recognition:

    python3 meterreader.py test-samples img

Process an entire folder of images:

    ./meterreader.py process-images img201911

To upload to influxdb:

    source .env

    (export variables - all 4)

    ./meterreader.py upload 2019

## Good resources

### Installing OpenCV (2019 new mac Edition)

### KNN

- https://towardsdatascience.com/scanned-digits-recognition-using-k-nearest-neighbor-k-nn-d1a1528f0dea

- https://stackoverflow.com/questions/9413216/simple-digit-recognition-ocr-in-opencv-python

## Notes

### Version of Jan 11 2019

SVM recognition:

46 recognized out of 79 images => 58%

516 recognized out of 553 digits => 93%

0: 25/26 - 96%

1: 183/183 - 100%

2: 22/36 - 61%

3: 31/33 - 94%

4: 32/34 - 94%

5: 86/88 - 98%

6: 49/50 - 98%

7: 19/26 - 73%

8: 35/42 - 83%

9: 34/35 - 97%

KNN recognition:

26 recognized out of 78 images => 33%

467 recognized out of 546 digits => 86%

0: 22/26 - 85%

1: 178/181 - 98%

2: 22/36 - 61%

3: 21/33 - 64%

4: 32/34 - 94%

5: 77/86 - 90%

6: 37/50 - 74%

7: 18/26 - 69%

8: 27/40 - 68%

9: 33/34 - 97%

### Version of Jan 9 2019

    ./meterreader.py train-samples img

    0: 36 - 6%

    1: 190 - 30%

    2: 46 - 7%

    3: 40 - 6%

    4: 43 - 7%

    5: 94 - 15%

    6: 56 - 9%

    7: 29 - 5%

    8: 54 - 8%

    9: 52 - 8%

    SVM Model saved!

    # Testing with only 6 digits.

    ./meterreader.py test-samples img

    71 recognized out of 83 images => 86%

    389 recognized out of 415 digits => 94%

    0: 13/14 - 93%

    1: 176/183 - 96%

    2: 17/18 - 94%

    3: 13/15 - 87%

    4: 19/20 - 95%

    5: 73/79 - 92%

    6: 37/39 - 95%

    7: 9/11 - 82%

    8: 16/19 - 84%

    9: 16/17 - 94%

### Version of Nov 17 2018

    ./meterreader.py train-samples img/

    0: 21 - 5%

    1: 121 - 29%

    2: 28 - 7%

    3: 29 - 7%

    4: 26 - 6%

    5: 60 - 14%

    6: 39 - 9%

    7: 22 - 5%

    8: 31 - 7%

    9: 39 - 9%

    SVM Model saved!

./meterreader.py test-samples img

13 recognized out of 54 images => 24%

357 recognized out of 424 digits => 84%