We have had some reports of low recognition rates. The recognition rate is the percentage of cards that the CardBot is confident have been recognised correctly - for example, if you load in 1,000 to be processed and 50 fail to be recognised, your recognition rate is 95%. We expect recognition rates to be above 90% (we can't guarantee a 100% recognition rate as cards can fail to be recognised due to reasons outside our control - damaged cards, upside down cards etc).

You can improve your CardBot's recognition rate by adjusting the color and brightness of the LED strips in the gantry head. You can edit these settings on the Advanced Configuration page of your Admin Interface, under the LED heading.

Before adjusting these settings, ensure your CardBot is placed in an ideal location (ie. not directly under a light source or facing a window with the sun) as this will help keep the outside lighting consistent and flat.

Adjusting the LED settings is very simple but can be time-consuming as you will only be changing one variable at a time. If you have any questions about tuning or would like a second opinion on any image logs, please send us a message through the Live Chat, or reach out via email at contact@cardcastle.co, and we can organise a time for a support call.

The steps to adjust your LED settings are as follows:

  1. Login in to the Admin Interface for your CardBot and go to the Settings page.

  2. Scroll down and tick the box in the 'Image Log' section and then click save - this will enable the CardBot to store up to 1,000 images of scanned cards. You will be using these images as a visual reference for changing the settings.

  3. Go back to the Home page and follow the URL that will lead to your CardBot's Control Panel.

  4. Run a scan cycle with ~20 cards of varying colors, noting the recognition rate.

  5. Go back to the Image Logging section as in Step 2 and download the .zip file.

  6. Unzip the file, then open the JPG images and analyse for glare, consistent lighting and color contrast - a guide on image quality is below.

  7. Go back to the Admin Interface and then go the Advanced Configurations page.

  8. Scroll down past the warning image and click on the right-most tab named 'LEDs' - this will show the LED settings for each phase of the CardBot's actions.

  9. Go to the second option down titled 'capturing' - this is only section that will affect your recognition rate so only change these settings.

  10. Depending on what you see in the images in Step 6, use the slider to adjust the brightness to a number that will allow for more even lighting - if the images have too much glare, lower the brightness, or if the images are too dark, increase the brightness.

  11. Scroll up and click the 'Save and Apply' button. Then repeat Steps 4-10 until you find the brightness setting that gives the best recognition rate.

  12. Once you have found the optimum brightness setting, go back to the 'capturing' section under the LED tab of the Advanced Configuration page and click on the box next to the 'color:' - this will open a panel where you can set the color of the LEDs as they light up for this stage of the scan cycle.

  13. Adjust the color by changing the RGB values at the bottom of the color panel - we recommend starting with the RGB values for Warm White (253, 244, 220).

  14. Once the color has been changed, click 'Save and Apply' button and repeat Steps 4-6, 12 and 13 until you see minimal glare, good lighting, and close to true colours - refer to the image quality guide below .

  15. Once the recognition rate is above ~85%, note down the values for brightness and RGB and run through a set of 1,000 cards to get a more accurate value for your recognition rate.

Here are some examples of images and how they can be improved:

Example 1: Not enough brightness

This CardBot's brightness is way too low, and as a result the card is not well-lit and evenly illuminated. The color doesn't actually need to be updated because while not lit up, the art isn't overly saturated and shows the colors accurately. There is also no glare, which is usually a consequence of color.

Example 2: The color needs to be adjusted

This image is a good example of the glare caused by the CardBot's LEDs being too bright a color- the outline of the card is well defined and evenly distributed while the symbols and words are clear, so the brightness is at a reasonable level. The glare is caused by the LEDs being too white, so the solution for this is to use a warmer white- this can often be greatly reduced by going to a Warm White (RGB is 253,244,220).

Example 3: An example of good conditions

This card is well lit and the borders are distinct and there is little to no washing out of color - the colors are in proportion and not too saturated, so the CardBot's machine vision can recognize this accurately!

Did this answer your question?