Try to detect the name based on the size of the text

[teolemon]

A simplistic assumption: the name of the product should be the largest text on the front of the product.
Based on this, we could compute a ratio between the area of the bounding boxes (width by height) and the amount of letters inside it .
Based on this ratio, we could have candidates for the product name

[CloCkWeRX]

I was thinking a good pipeline also could be:

• get all text from front image
• look for brand candidates from barcode group
• remove labels ("fat free")
• remove quantity (360g/similar regexp)
• extract remaining sentences
• weight "x with y" or "x & y" heavily (perhaps a good list of title patterns)

... and provide these as autocomplete suggestions for the name field.

These steps might help (particularly brands) filter put false positives