V2M4/rendering_video.py at main · WindVChen/V2M4 · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
import os
import imageio
import torch
import numpy as np
from v2m4_trellis.utils.render_utils import render_frames
from natsort import ns, natsorted
import trimesh
import argparse
import dill
from v2m4_trellis.representations.mesh import MeshExtractResult
import utils3d.torch as utils3d


def parse_args():
    parser = argparse.ArgumentParser(description='Trellis Benchmark')
    parser.add_argument('--baseline', action='store_true', help='Run the baseline model')
    parser.add_argument('--normal', action='store_true', help='Run the normal model')
    parser.add_argument('--interpolate', type=int, default=1, help='Interpolation steps between frames')
    parser.add_argument('--result_path', type=str, default='results', help='Path to the results folder')
    return parser.parse_args()


def get_folder_size(folder):
    """Returns the number of image files in the given folder."""
    return len([f for f in os.listdir(folder) if os.path.isfile(os.path.join(folder, f))])


if __name__ == "__main__":
    args = parse_args()
    root = args.result_path
    animations = [os.path.join(root, dir) for dir in os.listdir(root) if os.path.isdir(os.path.join(root, dir))]

    animations = natsorted(animations, alg=ns.PATH)

    assigned_animations = animations

    for animation in assigned_animations:
        # create folder output_path + "/" + "output_final_rendering_images"
        # create folder output_path + "/" + "output_final_rendering_images_baseline"
        os.makedirs(animation + "/" + "output_final_rendering_images", exist_ok=True)
        os.makedirs(animation + "/" + "output_final_rendering_images_baseline", exist_ok=True)

        renderings = []
        renderings_normal = []
        renderings_baseline = []
        renderings_baseline_normal = []

        renderings_interpolated = []
        renderings_interpolated_normal = []

        source_path = animation
        output_path = animation

        print("/n/n ============= Start processing: ", animation, " =============/n")

        glbs_list = os.listdir(source_path)
        # only keep file with suffix "_baseline_sample.glb"
        glbs_list = [glb for glb in glbs_list if "_texture_consistency_sample.glb" in glb]
        glbs_list = natsorted(glbs_list, alg=ns.PATH)

        if args.baseline:
            glbs_list_baseline = os.listdir(source_path)
            # only keep file with suffix "_baseline_sample.glb"
            glbs_list_baseline = [glb for glb in glbs_list_baseline if "_baseline_sample.glb" in glb]
            glbs_list_baseline = natsorted(glbs_list_baseline, alg=ns.PATH)
        else:
            glbs_list_baseline = glbs_list

        with open(output_path + "/extrinsics_list.pkl", 'rb') as f:
            extrinsics_list = dill.load(f)
            extr = extrinsics_list[0]

        last_mesh = None

        for ind, (glb, glb_baseline) in enumerate(zip(glbs_list, glbs_list_baseline)):
            mesh_file = source_path + "/" + glb
            mesh = trimesh.load(mesh_file, process=False)
            mesh = mesh.geometry["geometry_0"]
            mesh.vertices = mesh.vertices @ np.array([[1, 0, 0], [0, 0, 1], [0, -1, 0]])

            mesh_repre = MeshExtractResult(
                vertices=torch.tensor(mesh.vertices, dtype=torch.float32).cuda(),
                faces=torch.tensor(mesh.faces, dtype=torch.int64).cuda(),
                vertex_attrs=torch.tensor(mesh.visual.to_color().vertex_colors[..., :3], dtype=torch.float32).repeat(1, 2).cuda() / 255,
                res=512, texture=torch.tensor(np.array(mesh.visual.material.baseColorTexture), dtype=torch.float32).flip(0).cuda() / 255,
                uv=torch.tensor(mesh.visual.uv, dtype=torch.float32).cuda()
            )

            fovs = torch.deg2rad(torch.tensor(40, dtype=torch.float32)).cuda()
            intrinsics = [utils3d.intrinsics_from_fov_xy(fovs, fovs)]
            extrinsics = [torch.eye(4).cuda()]

            if (ind % args.interpolate == 0) and last_mesh is not None:
                # interpolation mesh
                for id in range(args.interpolate - 1):
                    temp_mesh_repre = MeshExtractResult(
                        vertices=last_mesh.vertices * (1 - id / args.interpolate) + mesh_repre.vertices * (id / args.interpolate),
                        faces=mesh_repre.faces,
                        vertex_attrs=torch.tensor(mesh.visual.to_color().vertex_colors[..., :3], dtype=torch.float32).repeat(1, 2).cuda() / 255,
                        res=512, texture=torch.tensor(np.array(mesh.visual.material.baseColorTexture), dtype=torch.float32).flip(0).cuda() / 255,
                        uv=torch.tensor(mesh.visual.uv, dtype=torch.float32).cuda()
                    )

                    img = render_frames(temp_mesh_repre, extrinsics, intrinsics, {'resolution': 1024, 'bg_color': (1, 1, 1)})['texture'][0]
                    renderings_interpolated.append(img)

                    if args.normal:
                        img_normal = render_frames(temp_mesh_repre, extrinsics, intrinsics, {'resolution': 1024, 'bg_color': (1, 1, 1)})['normal'][0]
                        renderings_interpolated_normal.append(img_normal)

            img = render_frames(mesh_repre, extrinsics, intrinsics, {'resolution': 1024, 'bg_color': (1, 1, 1)})['texture'][0]

            renderings.append(img)
            renderings_interpolated.append(img)

            if args.normal:
                img_normal = render_frames(mesh_repre, extrinsics, intrinsics, {'resolution': 1024, 'bg_color': (1, 1, 1)})['normal'][0]
                renderings_normal.append(img_normal)
                renderings_interpolated_normal.append(img_normal)

            imageio.imwrite(output_path + "/" + "output_final_rendering_images" + "/" + f"{ind:04d}.png", img)

            last_mesh = mesh_repre.deepcopy()

            "======== Re-canonicalization of the Mesh and Gaussian (*exclude RF*) ========"

            if args.baseline:
                mesh_file_baseline = source_path + "/" + glb_baseline
                mesh_baseline = trimesh.load(mesh_file_baseline, process=False)
                mesh_baseline = mesh_baseline.geometry["geometry_0"]
                mesh_baseline.vertices = mesh_baseline.vertices @ np.array([[1, 0, 0], [0, 0, 1], [0, -1, 0]])

                vertices = torch.from_numpy(mesh_baseline.vertices).unsqueeze(0).float().cuda()

                vertices_homo = torch.cat([vertices, torch.ones_like(vertices[..., :1])], dim=-1)
                vertices_camera = torch.bmm(vertices_homo, extr.transpose(-1, -2)).squeeze()

                mesh_baseline.vertices = vertices_camera[:, :3].cpu().numpy()

                mesh_repre_baseline = MeshExtractResult(
                    vertices=torch.tensor(mesh_baseline.vertices, dtype=torch.float32).cuda(),
                    faces=torch.tensor(mesh_baseline.faces, dtype=torch.int64).cuda(),
                    vertex_attrs=torch.tensor(mesh_baseline.visual.to_color().vertex_colors[..., :3], dtype=torch.float32).repeat(1, 2).cuda() / 255,
                    res=512, texture=torch.tensor(np.array(mesh_baseline.visual.material.baseColorTexture), dtype=torch.float32).flip(0).cuda() / 255,
                    uv=torch.tensor(mesh_baseline.visual.uv, dtype=torch.float32).cuda()
                )

                img_baseline = render_frames(mesh_repre_baseline, extrinsics, intrinsics, {'resolution': 1024, 'bg_color': (1, 1, 1)})['texture'][0]
                renderings_baseline.append(img_baseline)
                imageio.imwrite(output_path + "/" + "output_final_rendering_images_baseline" + "/" + f"{ind:04d}.png", img_baseline)

                if args.normal:
                    img_baseline_normal = render_frames(mesh_repre_baseline, extrinsics, intrinsics, {'resolution': 1024, 'bg_color': (1, 1, 1)})['normal'][0]
                    renderings_baseline_normal.append(img_baseline_normal)

        imageio.mimsave(output_path + "/" + "output_final_rendering_video.mp4", renderings, fps=30)
        imageio.mimsave(output_path + "/" + f"output_final_rendering_video_interpolated_{args.interpolate}.mp4", renderings_interpolated, fps=30)

        if args.normal:
            imageio.mimsave(output_path + "/" + "output_final_rendering_video_normal.mp4", renderings_normal, fps=30)

        if args.baseline:
            imageio.mimsave(output_path + "/" + "output_final_rendering_video_baseline.mp4", renderings_baseline, fps=30)

            if args.normal:
                imageio.mimsave(output_path + "/" + "output_final_rendering_video_baseline_normal.mp4", renderings_baseline_normal, fps=30)

        print("/n/n ============= Finish processing: ", animation, " =============/n")