Wenqi Li

> Hello everyone, > > I am new here and would like to contribute to implementing some (or all) of the evaluation metrics on @csudre's list. Did anyone start working...

today we discussed the requirement of FROC curves for lesion based evaluation

thank you, please feel free to create PRs (https://github.com/Project-MONAI/MONAI/blob/dev/CONTRIBUTING.md), preferably implementing this using only pytorch and monai APIs would make it easy to maintain for the core dev team.

Thanks and agree that adding an option will allow for different use cases, one of the use cases is to achieve fast but less precise results compared with the non-lazy......

I think we just use the `option` as you suggested to allow for the flexibility in general, the user can choose what to use for their own pipelines. it'll be...

sure, currently the main cases that I'm aware of are: - https://github.com/Project-MONAI/research-contributions/blob/18bb1d859fa65ad2221d2a6b6771e82b8e81ce89/auto3dseg/algorithm_templates/segresnet/scripts/segmenter.py#L200 - https://github.com/Project-MONAI/research-contributions/blob/main/auto3dseg/algorithm_templates/swinunetr/configs/transforms_train.yaml - https://github.com/Project-MONAI/research-contributions/blob/main/auto3dseg/algorithm_templates/dints/configs/transforms_train.yaml - https://github.com/Project-MONAI/tutorials/blob/main/modules/lazy_resampling_benchmark.ipynb - https://github.com/Project-MONAI/tutorials/blob/main/modules/lazy_resampling_compose.ipynb - https://github.com/Project-MONAI/tutorials/blob/main/modules/lazy_resampling_functional.ipynb - https://github.com/Project-MONAI/model-zoo/pull/397

> I've modified it to say 'will produce incorrect output'. I think I understand your point, but my point is that, there are 200+ transforms in monai, even if it's...

also, we'll never achieve exactly the same result by design, like the padding modes, output data types (quantisations), and interpolation modes. In most use cases, there's a trade-off for speed/interpolation...

On the dev branch there is no reordering of the user-defined transforms with or without lazy resampling, you can turn on the verbose=True to confirm this. So now I'm not...

The default behaviour is no lazy resampling, and the user should read the tutorials and examples before considering enabling lazy resampling. The definition of 'breaking their pipelines' is unclear here....