File size: 6,396 Bytes
0d1388f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import os
import sys

import argparse
import time
import imageio
import numpy as np
from PIL import Image

sys.path.insert(0, './hy3dshape')
os.environ['TORCH_CUDA_ARCH_LIST'] = '9.0'
os.environ['ATTN_BACKEND'] = 'xformers'
os.environ['SPCONV_ALGO'] = 'native'

from trellis.pipelines import NeARImageToRelightable3DPipeline
from hy3dshape.pipelines import Hunyuan3DDiTFlowMatchingPipeline


def main():
    parser = argparse.ArgumentParser(description='NeAR: from image or SLaT to relightable 3D')
    parser.add_argument('--checkpoint', type=str, default='checkpoints',
                        help='Pipeline weights directory (contains pipeline.yaml)')
    parser.add_argument('--image', type=str, default="assets/example_image/T.png",
                        help='Input image path (one of --slat or --image)')
    parser.add_argument('--slat', type=str, default=None,
                        help='SLaT .npz path (one of --image or --slat)')
    parser.add_argument('--hdri', type=str, default="assets/hdris/studio_small_03_1k.exr",
                        help='HDRI .exr path')
    parser.add_argument('--out_dir', type=str, default='relight_out',
                        help='Output directory')
    parser.add_argument('--yaw', type=float, default=0.0, help='View yaw (degrees) ')
    parser.add_argument('--pitch', type=float, default=0.0, help='View pitch (degrees)')
    parser.add_argument('--hdri_rot', type=float, default=0.0,
                        help='HDRI rotation angle (degrees)')
    parser.add_argument('--video_frames', type=int, default=40,
                        help='Render additional spiral camera path video frames')
    parser.add_argument('--seed', type=int, default=42, help='Random seed')
    parser.add_argument('--save_slat', type=str, default=None,
                        help='When generating from image, save SLaT to this .npz path')
    parser.add_argument('--no_cuda', action='store_true', help='Do not use CUDA')
    args = parser.parse_args()

    if (args.image is None) == (args.slat is None):
        parser.error('Please specify --image or --slat, one of them')
    from_image = args.image is not None

    os.makedirs(args.out_dir, exist_ok=True)
    device = 'cuda' if not args.no_cuda else 'cpu'

    total_t0 = time.perf_counter()
    t0 = time.perf_counter()
    hyshape_model_id = 'tencent/Hunyuan3D-2.1'
    hyshape_pipe = Hunyuan3DDiTFlowMatchingPipeline.from_pretrained(hyshape_model_id)
    hyshape_pipe.to(device)
    pipeline = NeARImageToRelightable3DPipeline.from_pretrained(args.checkpoint)
    pipeline.to(device)
    print(f"  [OK] Pipeline loaded, +{time.perf_counter() - t0:.1f}s")

    if from_image:
        image = Image.open(args.image).convert('RGB')
        image_prep = pipeline.preprocess_image(image)

        t0 = time.perf_counter()
        mesh = hyshape_pipe(image=image_prep)[0]
        mesh_path = os.path.join(args.out_dir, 'initial_3d_shape.glb')
        mesh.export(mesh_path)
        print(f"  [OK] Geometry mesh generated, +{time.perf_counter() - t0:.1f}s, saved to {mesh_path}")

        t0 = time.perf_counter()
        slat = pipeline.run_with_shape(
            image_prep,
            mesh,
            seed=args.seed,
            preprocess_image=False,
        )
        print(f"  [OK] Image → SLaT generated, +{time.perf_counter() - t0:.1f}s")
        if args.save_slat:
            np.savez(
                args.save_slat,
                feats=slat.feats.cpu().numpy(),
                coords=slat.coords.cpu().numpy(),
            )
            print(f"  [OK] Saved SLaT to: {args.save_slat}")
    else:
        t0 = time.perf_counter()
        slat = pipeline.load_slat(args.slat)
        print(f"  [OK] Loaded SLaT, +{time.perf_counter() - t0:.1f}s")

    t0 = time.perf_counter()
    hdri_np = pipeline.load_hdri(args.hdri)
    print(f"  [OK] Loaded HDRI, +{time.perf_counter() - t0:.1f}s")

    t0 = time.perf_counter()

    pipeline.renderer.ssaa = 1
    pipeline.renderer.resolution = 1024

    views = pipeline.render_view(
        slat, hdri_np,
        yaw_deg=args.yaw, pitch_deg=args.pitch,
        fov=40.0, radius=2.0, hdri_rot_deg=args.hdri_rot,
        resolution=512
    )

    color_path = os.path.join(args.out_dir, 'relight_color.png')
    base_color_path = os.path.join(args.out_dir, 'base_color.png')
    metallic_path = os.path.join(args.out_dir, 'metallic.png')
    roughness_path = os.path.join(args.out_dir, 'roughness.png')
    shadow_path = os.path.join(args.out_dir, 'shadow.png')
    views['color'].save(color_path)
    views['base_color'].save(base_color_path)
    views['metallic'].save(metallic_path)
    views['roughness'].save(roughness_path)
    views['shadow'].save(shadow_path)

    print(f"  [OK] Single view rendering completed, +{time.perf_counter() - t0:.1f}s, saved to {color_path}, {base_color_path}, {metallic_path}, {roughness_path}, {shadow_path}")

    if args.video_frames > 0:
        t0 = time.perf_counter()
        frames = pipeline.render_camera_path_video(
            slat, hdri_np, num_views=args.video_frames,
            fov=40.0, radius=2.0, hdri_rot_deg=args.hdri_rot,
            verbose=True, full_video=True, shadow_video=True
        )
        video_path = os.path.join(args.out_dir, 'relight_camera_path.mp4')
        imageio.mimsave(video_path, frames, fps=24)
        print(f"  [OK] Camera path video completed, +{time.perf_counter() - t0:.1f}s, saved to {video_path}")
    print("Done.")

    if args.video_frames > 0:
        t0 = time.perf_counter()
        hdri_roll_frames, render_frames = pipeline.render_hdri_rotation_video(
            slat, hdri_np, num_frames=args.video_frames,
            yaw_deg=args.yaw, pitch_deg=args.pitch,
            fov=40.0, radius=2.0,
            verbose=True, full_video=True, shadow_video=True
        )
        hdri_video_path = os.path.join(args.out_dir, 'hdri_roll.mp4')
        render_video_path = os.path.join(args.out_dir, 'relight_hdri_rotation.mp4')
        imageio.mimsave(hdri_video_path, hdri_roll_frames, fps=24)
        imageio.mimsave(render_video_path, render_frames, fps=24)
        print(
            f"  [OK] HDRI rotation videos completed, +{time.perf_counter() - t0:.1f}s, "
            f"saved to {hdri_video_path} and {render_video_path}"
        )

    print(f"  [OK] Total time: {time.perf_counter() - total_t0:.1f}s")


if __name__ == '__main__':
    main()