search

devmaxxing / videocr-PaddleOCR

Extract hardcoded subtitles from videos using machine learning
MIT License
138 stars 21 forks source link

Google colab Notebook problem #22

Open myway24 opened 1 month ago

myway24 commented 1 month ago

error: OpenCV(4.6.0) /io/opencv/modules/imgproc/src/color.cpp:182: error: (-215:Assertion failed) !_src.empty() in function 'cvtColor'

When I click OCR parameters, This error happens. What should I do?

devmaxxing commented 1 month ago

I think I might have seen this before when the start time is at 0 or the end time is at the last possible time stamp. Does adjusting the time range fix this?

myway24 commented 1 month ago

I don't know why but it works now anyway(same file and settings) even if the start time begins at 0:00, did you by any chance update the colab codes? thanks for your work,this ocr based srt is extremely useful for language learner like me especially when you can't find out popular Chinese drama subtitles which are to be used for getting time stamps in my case because language apps usually specify repeating section using the time stamps attached to txt as you already know, less accurate txt is not a problem, only if you get Time stamps, accurate txt is already hardcoded on video,and I would suggest one more feature, is it possible to generate srt using ocr and whisper ai at the same time?With more data, you can produce more accurate subtitles, perhaps many people (including me)want it to operate on colab due to limited pc resources,especially expensive graphic card! Keep up the good work! Ps whisper ai is relatively poor at generating Chinese srt,is there any other reliable open source speech to txt engine or API in china? if any, please make it also available on colab,this too is for language learning, thank you.