TABLE OF CONTENTS
- What is the READ API?
- Is the READ API secure?
- What is the accuracy rate of the READ API?
- How do I get more API calls?
- What can I submit to the READ API?
What is the READ API?
Vidado READ API is our cloud-based OCR engine that supports zones (not pages) -- each image must be smaller than 100,000 pixels, which is 1.1 inch² or 7.1 cm² at 300 dpi. The script will dynamically shrink images that are larger than 100,000 pixels. Vidado supports jpeg, jpg, gif, png. The script uses png because it supports B&W, grey and color losslessly.
Only horizontal text can be read, not vertical. There is no deskew, but the engine was trained on samples with ±5° skew.
The engine is a Deep Neural Network trained on 1,000,000,000 hand-checked field samples in English. You can expect excellent results on Latin languages without accents - eg Spanish, Dutch, Italian, Swahili, Indonesian. (It will not return accents). It was trained to read words on forms and hence it ignores labels. In the example below "christina" is found and "witness signature" is ignored.
As mentioned in the results above, the READ API only returns results in lower case. Learn more about the READ API here.
Is the READ API secure?
Vidado does not keep any images or results after processing.
What is the accuracy rate of the READ API?
Vidado claims over 90% field level accuracy or better than a human.
How do I get more API calls?
The trial license is limited to 1 call per second. Please contact your Sales representative for more API calls, or feel free to reach out to firstname.lastname@example.org.
What can I submit to the READ API?
You are permitted to submit jpeg, jpg, gif, or png images that are smaller than 100,000 pixels, which is 1.1 inch² or 7.1 cm² at 300 dpi. The script will dynamically shrink images that are larger than 100,000 pixels. The script uses png because it supports B&W, grey and color losslessly. Image cleanup is not required - smudges, lines, boxes, shadings are all ignored.