Clarification on Metric Information Output in Pretrained Weights

[wjy5446]

Hello,

I have a question regarding the pretrained model named MASt3R_ViTLarge_BaseDecoder_512_catmlpdpt_metric available in this repository.

Does this model provide metric information during inference?
I noticed that the model name includes the term metric, and I was wondering if this implies that it outputs metric information as part of its inference process.

Comparison with Dust3r Model:
In contrast, the Dust3r model does not seem to include metric in its name. This leads me to believe that the metric term in MASt3R pretrained weights might signify additional capabilities or outputs related to metrics.

Clarification Request:
If my assumption is incorrect, could you clarify the meaning of metric in the context of the pretrained weights?

Thank you for your time and help in addressing my question!