Re-train a reconstructed scene, adding extra images

[vrahnos3]

Hello,
I know that with COLMAP and HLOC is possible to register new images in an existing reconstrurtion.
My question is:
If I add some new images, does it need to train from the beggining the nerfacto model?
Is there anyway to continue the training from the last checkpoint but with new images or do I need to train a new model from the beginning?

Thanks in advance

[Ben-Mack]

I'm also highly interested in this workflow, which if possible will help reduce a lot of time to retrain from ground up.
Does anyone know any workflow that make this possible?