If no text has been recognized or if text recognition has not worked well, check the following.
Go through the Raster Text Quality Checklist
Scan2CAD’s OCR text recognition will only recognize raster text where the following conditions are met:
Use the Raster Text Quality Checklist to ensure that your text meets these conditions.
Is the text very small?
Scan2CAD is unlikely to recognize text that is smaller than five pixels in size, or even text that is slightly larger than this where letters are complex. This is because where text is very small there are simply not enough pixels in each letter to adequately represent the letter’s shape so that it can be recognized.
Check that the OCR settings that are set are suitable for the text you are converting.
It is important to set the Character Size Limits correctly. Scan2CAD will only recognize an item as text if it is smaller than the Maximum Character Size Limit but larger than the Minimum Character Size Limit.
Do not set the Maximum Character Size Limit to a very large value in the hope of recognizing all the text on the image. If you do this, Scan2CAD is likely to miss text on the image even though it is smaller than the Maximum Limit. Set the Maximum Character Size Limit to a value slightly larger than the height of most of the text on the image. Do this using OCR Menu > Char Size.
Make sure that the Character Rotation matches the orientation of the text on the image.
If Scan2CAD is splitting text strings that should not be split or is joining text strings that should be separate, try editing the Character Space Limit. If increasing the Character Space Limit does not join two text strings, try doubling the Maximum Character Size Limit as well.
If the recognized text looks like this it is because Display Mode in OCR Menu > Settings is set to Chars.
Press the ? key on the keyboard. If nothing appears click on OCR Menu > Build Strings to join the individual characters together to make text strings.
Does the image contain areas of numbers?
Scan2CAD recognizes text by means of technology called neural networks. Different neural networks must be used for optimal recognition of different fonts or character sets.
If you have an image or an area of an image that contains numbers only, for example an area of dimensions, it is a good idea to convert that area of the image using a neural network that has been trained to recognize numbers only. This means that there is less likely to be confusion between similar numeric and non-numeric characters such as 8 and B.
Scan2CAD is supplied with a neural network called Number.Net that has been trained to recognize numbers and symbols only.
To use Number.Net, check the User option in OCR Menu > Settings and select Number.Net from the drop down list.
To use Number.Net, check the User option in OCR Menu > Settings.
Is a Pick Color selected?
If a Pick Color is selected, Scan2CAD only recognizes Pick Colored text.
Before converting an image with a Pick Color selected you must reduce the colors on the raster image or conversion will not work as you expect.
Are Layer Colors activated?
When Color Layers are activated Scan2CAD only recognizes Layer Colored text.
Before converting an image with Layer Colors activated you must reduce the colors on the raster image or conversion will not work as you expect.
Is the text white?
Scan2CAD does not recognize white text because it assumes that the paper is white and that the text is non-white.
Did you convert the text using Run Menu > All?
If you converted the image using Run Menu > All () and no text has been recognized, go to the Type Menu and see which option is selected.
If Outline or Scanline are selected, Scan2CAD won’t recognize text when you convert the image using Run Menu > All. If you have selected Outline or Scanline you need to convert the text first using Run Menu > OCR (), then convert the rest of the image using Run Menu > Vectorization ().
Are there question marks in the text?
Is text that should be the same height at different heights?
Make vector text the same height using OCR Menu > Unify String Heights.
“The number of raster objects exceeds 32000” error message
If you get this error message you probably have a very speckled image. Try the following:
Troubleshooting Text recognition
Have questions on this topic? Talk to us