How to train Foundation pose for custom dataset ?

[Manu752]

Hello,

I’ve been following the discussion here: FoundationPose estimation with custom object

Based on that thread, I trained RT-DETR for my custom object and now have solid detection results. The next step is to use FoundationPose for 6DoF pose estimation, but I’m unclear on the pipeline for:

Preparing custom object data in the required format.

Training or fine-tuning FoundationPose on a new object.

Using RT-DETR outputs as the detector stage for FoundationPose inference.

Could you clarify:

Is there an official workflow or example for training FoundationPose on a new object?