Can't reproduce the same results as neuralangelo.

[cduguet]

Hello,

We have succesfully captured a real scene of a saxophone with neuralangelo, and the results look decent. I am trying to reenact the same results with instant-angelo but so far I am failing. What should I do?

This is the dataset, and neuralangelo's result.
My configs in neuralangelo were:

_parent_: projects/neuralangelo/configs/base.yaml
data:
    readjust:
        center:
        - 0.0
        - 0.0
        - 0.0
        scale: 0.2
    root: /data/datasets/sax/fullres/saxophone-fullres-undistorted
    train:
        image_size:
        - 5936
        - 3956
    type: projects.neuralangelo.data
    val:
        image_size:
        - 450
        - 300
model:
    appear_embed:
        enabled: false
    object:
        sdf:
            encoding:
                coarse2fine:
                    init_active_level: 4
            mlp:
                inside_out: false

I am aware you have another repository to share data and results, but I am not sure how to contribute, hence this discussion.

Is there anything I'm missing on instant-angelo?

Thanks !

[xsl316]

May I ask what difficulties you encountered during the reproduction of Instant-angelo?