jsbroks / imantics

:camera: Reactive python package for managing, creating and visualizing different deep-learning image annotation formats

https://imantics.rtfd.io

MIT License

364 stars 68 forks source link

Handling large datasets #18

Open EtienneDavid opened 4 years ago

EtienneDavid commented 4 years ago

Hello,

I recently faced an issue with the tool while converting a very large dataset. The problem is that in the dataset object the image is loaded and save, for drawing purposes. But for very large dataset it means blowing away the RAM...

I don't know if having an option to disable the drawing for large dataset is something we want or not... What do you think ?

jsbroks commented 4 years ago

Going point. No need to load all the images if you only want to do annotation manipulation. Probably lazy loading it would be best

EtienneDavid commented 4 years ago

OK, i will think about it and probably propose a PR

EtienneDavid commented 4 years ago

20

Petros626 commented 2 years ago

@EtienneDavid the commits were to handel large datasets right? so i should do it on my own, the author did not update it?