Normalization During Inference

[violag12]

Thank you for your awesome work!

Shouldn't the point cloud extracted from the input mesh be normalized in the dataset.py code so that it matches the distribution of point clouds used during training? Should center_vertices and normalize_vertices_scale be called in dataset.py?

DeepMesh/sft/datasets/dataset.py

Line 49 in b5964ab

verts, faces = mesh.vertices,mesh.faces

DeepMesh/sft/datasets/data_utils.py

Lines 108 to 116 in b5964ab

	# Translate the vertices so that bounding box is centered at zero.
	vertices = center_vertices(vertices)
	if augment:
	vertices = augment_mesh(vertices, **augment_dict)
	# Scale the vertices so that the long diagonal of the bounding box is equal
	# to one.
	vertices = normalize_vertices_scale(vertices)

[guoshuzhen-jpg]

May I ask what format your input data is?ply?glb?