•1 min read•from Machine Learning
PaddleOCR (v3/v4/v5/v6) implemented in C++ with ncnn [P]
Hi,
About a year ago I shared my PaddleOCR implementation here. Since then I've made many improvements, and it now supports PP-OCR v3 through the latest v6 models.
The official Paddle C++ runtime has a lot of dependencies and is very complex to deploy. To keep things simple I use ncnn for inference, it's much lighter (and faster in my task), makes deployment easy.
Hope it's helpful to some of you, and feedback welcome!
[link] [comments]
Want to read more?
Check out the full article on the original site