Closed lambdaofgod closed 2 years ago
The example of model training in documentation is misleading, as it's not actually end-to-end - it trains detector and recognizer separately.
The example actually shows training of all parts of pipeline. The process isn't end-to-end though.
Thanks for raising this. I agree that the typical use of the term "end-to-end" in deep learning doesn't apply here and that the current wording is misleading. Would you be willing to file a PR to fix/change it?
The example of model training in documentation is misleading, as it's not actually end-to-end - it trains detector and recognizer separately.
The example actually shows training of all parts of pipeline. The process isn't end-to-end though.