Inference speed
#1
by johnlockejrr - opened
Not trolling or praising but anyway... I never encountered such speed in inference like PaddlePaddle/PaddleOCR has, is lightning fast!
Thanks for the kind words! Our PaddlePaddle/PP-OCRv6_tiny_det can run even faster, and you can even try deploying it on MCUs. Welcome to test it out on various edge devices!
And I wasn't very exact in my post above and the image. That is not just OCRv6_medium_det alone, is PP-DocLayoutV3 -> PP-OCRv6_medium_det, 570 pages in just a little over a minute!
Thanks for the support! I'll be sharing more deep dives and behind-the-scenes stories about Baidu OCR on my X account. Feel free to follow along here: https://x.com/slimcat0101