You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
By default, the mmocr/ocr.py script uses cuda as the device, as opposed to automatically detecting. All of the install and quick start docs suggest you run commands like python mmocr/ocr.py --det DB_r18 --recog CRNN demo/demo_text_ocr.jpg --show to test the system, but these do not work, and fail in an unexpected way that sends users off trying to fix they pyTorch installs.
The docs already show how to otherwise get the system running on a CPU. Could you provide CPU and CUDA examples for these test scripts, at least in the installation.
Thanks for the feedback. We will release a new inference implementation that will automatically detect the best device to use. I guess we don't need to revise the docs at this moment.
Branch
master branch https://mmocr.readthedocs.io/en/latest/
📚 The doc issue
By default, the
mmocr/ocr.py
script usescuda
as the device, as opposed to automatically detecting. All of the install and quick start docs suggest you run commands likepython mmocr/ocr.py --det DB_r18 --recog CRNN demo/demo_text_ocr.jpg --show
to test the system, but these do not work, and fail in an unexpected way that sends users off trying to fix they pyTorch installs.The docs already show how to otherwise get the system running on a CPU. Could you provide CPU and CUDA examples for these test scripts, at least in the installation.
Suggest a potential alternative/fix
Run the following in MMOCR’s directory:
or, for CPU only
The text was updated successfully, but these errors were encountered: