How can I configure inference to consider multiple images

[luisCartoGeo]

Hello, another question. How can I configure the inference to consider several images of the same scene, so that I can take advantage of the different views and angles? In the Hugging Face demo and the tests I have carried out, introducing the images by Loads or by pre-processing, adding intrinsic features generates separate inferences.
Since there is no specific parameter to tell the model to consider all images as from the same scene, if I specify the poses, would it process two (stereo) or more images as from the same scene?
I would appreciate your guidance on this, if possible.

[Nik-V9]

Hi, I'm a bit confused by the question.

I'm not sure I get what you mean by separate inferences. Do you mean that you are running the model.infer separately on each image from a scene?

By default, when you pass in a list of dictionaries to the model inference call, it will use the multi-view context to make the predictions.

Re specifying poses: The performance of MapAnything can be further improved if you provide poses as additional input.

[luisCartoGeo]

Hello, yes, apologies, I had misinterpreted the results. Although results are generated for each image (image, point clouds, poses, others), they are processed for the same scene, if entered correctly, a list of view dictionaries. Therefore, at the end, I can superimpose these results.