Hi! Nifty software. I ran into some major accuracy issues using any of the 5 built-in OCR options, though (on Bravely Default in JP, specifically); it regularly returns completely incorrect results even for basic kana-only words. I've seen similar issues with Tesseract before (which I'd used from command line in the past), and found a lot of labor-intensive image cleanup could often fix them, but it never really seemed very consistent.
I've never had similar issues with Google's Vision API, though. Whether from command line, phone, or browser, it kind of 'just works' for Kanji/kana recognition. I was wondering how difficult it would be to add Vision as an available OCR engine, and if @mathewthe2 has considered it before?
Hi! Nifty software. I ran into some major accuracy issues using any of the 5 built-in OCR options, though (on Bravely Default in JP, specifically); it regularly returns completely incorrect results even for basic kana-only words. I've seen similar issues with Tesseract before (which I'd used from command line in the past), and found a lot of labor-intensive image cleanup could often fix them, but it never really seemed very consistent.
I've never had similar issues with Google's Vision API, though. Whether from command line, phone, or browser, it kind of 'just works' for Kanji/kana recognition. I was wondering how difficult it would be to add Vision as an available OCR engine, and if @mathewthe2 has considered it before?