Open maxminstr opened 1 year ago
If you want to add new language, plz train your own langauge recognition mode. Once you get a Finnish model, I can add to this software and make support. Training Turtorial can be found at: https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.6/doc/doc_en/training_en.md
Hi, maxminstr
I just update the source code, and I found that you may try recognising Finnish by select Latin.
Thanks for the tip!
I suppose source video quality might have an impact as I'm testing with a captured VCR recording: I might later test first upscaling the source video.
When using English as detection language, vse properly puts two-row subtitles to two rows also in the resulting srt. However, with other detection languages (such as Latin) two-row subtitles are put on single row. Any way to fix this for Latin for example?
I was previously using pre-built Windows EXE, but now I used the latest source code.
However, when setting up the environment, I got this error:
ERROR: Cannot install -r requirements.txt (line 10), -r requirements.txt (line 13), -r requirements.txt (line 16), -r requirements.txt (line 17), -r requirements.txt (line 9) and numpy==1.20.0 because these package versions have conflicting dependencies.
The conflict is caused by:
The user requested numpy==1.20.0
imageio 2.13.3 depends on numpy
imgaug 0.4.0 depends on numpy>=1.15
matplotlib 3.5.1 depends on numpy>=1.17
opencv-python 4.5.4.60 depends on numpy>=1.17.3
paddlepaddle 2.2.2 depends on numpy<=1.19.3 and >=1.13; python_version >= "3.5" and platform_system == "Windows"
I was able to fix this by removing the version requirement from "paddlepaddle".
So far, I didn't notice any adverse effects.
Would it be possible to add support for Finnish?
If there is anything I can do to help you accomplish this, please let me know.