Changes

Jump to navigation Jump to search
184 bytes added ,  13:01, 28 June 2018
Line 355: Line 355:     
==== Performing OCR on a photo ====
 
==== Performing OCR on a photo ====
We start from a source image (defined by barcode and id) as selected for ingredients. We ask Product Opener to process it (process_image=1). Product Opener returns a JSON. Processing is done using Tesseract (3 on preprod). The result is often cripped with errors. Take steps to avoid direct copy pasting into the ingredients field by your users.
+
We start from a source image (defined by barcode and id) as selected for ingredients. We ask Product Opener to process it (process_image=1). Product Opener returns a JSON. Processing is done using Tesseract or Google Cloud Vision. The result is often cripped with errors with Tesseract, less with Google Cloud Vision. Take steps to ensure your users correct the output. Do send the corrected output using the ingredients WRITE API. If possible, please use your own OCR, this costs us money, and we're a non-profit with no budget.
    
<pre>Test server: https://world.openfoodfacts.net/cgi/ingredients.pl</pre>
 
<pre>Test server: https://world.openfoodfacts.net/cgi/ingredients.pl</pre>

Navigation menu