cvcg_utils.uv package

cvcg_utils.uv.img2mesh module

cvcg_utils.uv.img2mesh.make_masked_delaunay(mask)

mask: [H, W], bool

cvcg_utils.uv.img2mesh.make_pixel_indices(H, W)

cvcg_utils.uv.img2mesh.make_pixel_triangles(H, W)

generates triangles for an image of size (H, W):

pixel indices: [[0, 1, 2, …, W-1],

[W, W+1, …, 2W-1], … […, HW-1]]

cvcg_utils.uv.img2mesh.make_pixel_uvs(H, W, homogeneous)

generates a pixel uv coord image uv_img of shape [H, W, 2], where uv_img[i, j] = [i+0.5, j+0.5] is the center coordinates of pixel (i, j):

[[0.5,   0.5], [1.5,   0.5], [2.5,   0.5], ..., [W-1.5,   0.5],
 [0.5,   1.5], [1.5,   1.5], [2.5,   1.5], ..., [W-1.5,   1.5],
 ...
 [0.5, H-1.5], [1.5, H-1.5], [2.5, H-1.5], ..., [W-1.5, H-1.5]]

If homogeneous is True, the uv_img is padded with a constant value 1.

Parameters:

• H (int) – height of the image

• W (int) – width of the image

• homogeneous (bool) – whether to make uv_img homogeneous

Returns:

uv_img – uv coord image of shape [H, W, 2] or [H, W, 3] if homogeneous is True

Return type:

np.ndarray

cvcg_utils.uv.img2mesh.make_texture_uvs(H, W)

generates texture uv coord image

cvcg_utils.uv.posmap2normal module

class cvcg_utils.uv.posmap2normal.Img2Grad(*args: Any, **kwargs: Any)

Bases: Module

forward(img: torch.Tensor, mask: torch.Tensor) → torch.Tensor

img: [b, c, h, w] mask: [b, h, w], bool

class cvcg_utils.uv.posmap2normal.MaskedPosmap2Normal(*args: Any, **kwargs: Any)

Bases: Module

forward(posmap: torch.Tensor, normalize=True) → torch.Tensor

posmap: [b, 3, h, w] normal_map: [b, 3, h, w]

class cvcg_utils.uv.posmap2normal.Posmap2Normal(*args: Any, **kwargs: Any)

Bases: Module

forward(posmap: torch.Tensor, mask: torch.Tensor, normalize=True) → torch.Tensor

posmap: [b, 3, h, w] normal_map: [b, 3, h, w] mask: [b, h, w], bool flip_mask: Not supported yet

cvcg_utils.uv.test_rgb_xyz_to_mesh module

cvcg_utils.uv.test_rgb_xyz_to_mesh_delaunay module

cvcg_utils.uv.uv_module module

cvcg_utils.uv.uv_module.convert_edge_uv_to_uv_mesh(edge_uv: torch.Tensor)

edge_uv: [N_faces, 3, 2] tensor storing edge uv coords, compatible with PLY formats.

cvcg_utils.uv.uv_module.face_attr_to_uv(face_attrs, face_index_map, uv_mask)

face_attrs: [B, N, C] face_index_map: [H, W, 3] uv_mask: [H, W]

cvcg_utils.uv.uv_module.get_bary_coords_2d(query_coords: torch.Tensor, target_tris: torch.Tensor)

Assuming we already know the triangles of each query point, i.e., dim 0 of query_coords and target_triangles are in correspondence

query_coords: [Nv, 2 (uv coords)] target_tris: [Nv, 3 (vertices of the tri), 2 (uv coords)]

cvcg_utils.uv.uv_module.get_uv_assets(faces: torch.Tensor, uv_verts: torch.Tensor, uv_faces: torch.Tensor, uv_size: Tuple[int, int] | int, flip_v: bool = True) → Tuple[torch.Tensor, torch.Tensor]

get the following assets for uv-based utils:

vert_index_map: [H, W, 3] face_index_map: [H, W] bary_coords_map: [H, W, 3] uv_mask: [H, W]

cvcg_utils.uv.uv_module.get_uv_face_index_map(uv_verts: torch.Tensor, uv_faces: torch.Tensor, uv_size: Tuple[int, int] | int, flip_v: bool = True) → torch.Tensor

uv_verts: [Nv_uv, 2], uv-space vertices uv_faces: [Nf, 3] triangles uv_size: (H, W) or just int

out: [H, W], with -1 denoting no triangles

cvcg_utils.uv.uv_module.sample_texture_by_uv(uv: torch.Tensor, tex_img: torch.Tensor, padding_mode='border')

uv: [B, …, 2], (0, 0) is bottom left, (1, 1) is top right tex_img: [B, C, H, W]

out: [B, …, C]

cvcg_utils.uv.uv_module.vert_attr_to_uv(vert_attrs, vert_index_map, bary_coords_map, uv_mask)

vert_attrs: [B, N, C] vert_index_map: [H, W, 3] bary_coords_map: [H, W, 3] uv_mask: [H, W]

Module contents