Using cloud mask on subset imagery during K means classification

Hi there,
I am getting myself a bit confused by the cloud masks for S2 products. Basically I am trying to run a K means classification on a subset image. I understand that the cloud masks are available at the tile level but once you resample and subset these are no longer available. I have created a cloud mask band using the methodology outlined in the RUS webinar on NBR but I am now not sure how to add this as an RoI mask within the classification process. Do I need to convert this into a vector somehow?

I have tried using one of the cloud masks in the K means at the tile level but this didnt work as it applied the K means to the areas of cloud not the rest of the tile outwith the cloud, do I need to do some kind of inversion here?

Any help would be much appreciated.

cheers the now

The cloud masks are still in the product after subsetting and resampling, at least the scl masks, that include the cloud identification are still in the Mask Manager (I made the test with L2A).

I see is not the same with L1C…
You could apply: Resampling–> Subet–>Idepix
Idepix would identify many types of cloudy pixels and you can use them as ROIs in your K-means.