Simple3Diffusion/main.py at main · ailllist/Simple3Diffusion · 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 numpy as np
import open3d as o3d
from tqdm import tqdm

import torch
from torch.utils.data import DataLoader
from torch.optim.lr_scheduler import LambdaLR

from model import *
from dataset import ShapeNetCore

def get_linear_scheduler(optimizer, start_epoch, end_epoch, start_lr, end_lr):
    def lr_func(epoch):
        if epoch <= start_epoch:
            return 1.0
        elif epoch <= end_epoch:
            total = end_epoch - start_epoch
            delta = epoch - start_epoch
            frac = delta / total
            return (1-frac) * 1.0 + frac * (end_lr / start_lr)
        else:
            return end_lr / start_lr

    return LambdaLR(optimizer, lr_lambda=lr_func)


def get_data_iterator(iterable):
    """Allows training with DataLoaders in a single infinite loop:
        for i, data in enumerate(inf_generator(train_loader)):
    """
    iterator = iterable.__iter__()
    while True:
        try:
            yield iterator.__next__()
        except StopIteration:
            iterator = iterable.__iter__()

path = './data/shapenet.hdf5'
cate = "airplane"
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
raw = False  # View results for all time steps (including saving)

train_dataset = ShapeNetCore(
    path=path,
    cate=cate,
    split='train'
)
val_dataset = ShapeNetCore(
    path=path,
    cate=cate,
    split='val'
)

model = DiffusionPoint(
    SimpleNet(),
    VarianceSchedule()
).to(device)

train_iter = get_data_iterator(DataLoader(
    train_dataset,
    batch_size=2,
    shuffle=True
))

val_iter = get_data_iterator(DataLoader(
    val_dataset,
    batch_size=1,
    shuffle=True
))

optimizer = torch.optim.Adam(model.parameters(), lr=2e-3, weight_decay=0)
schedular = get_linear_scheduler(optimizer, 200000, 400000, 2e-3, 1e-4)

def save_pcds(pcds: dict):
    vis = o3d.visualization.Visualizer()
    vis.create_window()
    for t, pcl in pcds.items():
        pcl = pcl.cpu().numpy()

        pcd = o3d.geometry.PointCloud()
        pcd.points = o3d.utility.Vector3dVector(pcl[0])

        color = np.array([[1, 0, 0]] * pcl.shape[0])
        pcd.colors = o3d.utility.Vector3dVector(color)

        vis.add_geometry(pcd)
        vis.poll_events()
        vis.update_renderer()

        view_control = vis.get_view_control()
        view_control.set_up([0, 1, 0])
        view_control.set_front([1, 1, 1])
        view_control.set_lookat([0, 0, 0])
        view_control.set_zoom(0.5)
        vis.poll_events()
        vis.update_renderer()
        if os.path.exists("imgs"):
            pass
        else:
            os.mkdir("imgs")

        img_pth = f"imgs/{t}.png"
        vis.capture_screen_image(img_pth)
        vis.clear_geometries()
        print(f"saved image at {t}")

    vis.destroy_window()

def train(it):
    x = next(train_iter)
    x = x.to(device)
    optimizer.zero_grad()
    model.train()

    loss = model.get_loss(x)
    print(it, ":", loss)
    loss.backward()
    optimizer.step()
    schedular.step()

def test(it):  # Visualization
    tgt = next(val_iter)
    tgt = tgt[0]
    num_points, coord = tgt.size()
    tgt = np.asarray(tgt)
    tgt = tgt - np.array([0, 2, 0])

    with torch.no_grad():
        res = model.sample(1, num_points, coord, device, ret_traj=raw)
        if raw:
            save_pcds(res)
            res = res[0]
            points_numpy = res[0].cpu().numpy()
        else:
            points_numpy = res[0].cpu().numpy()

        pcd1 = o3d.geometry.PointCloud()
        pcd1.points = o3d.utility.Vector3dVector(points_numpy)

        color = np.array([[1, 0, 0]] * points_numpy.shape[0])
        pcd1.colors = o3d.utility.Vector3dVector(color)

        pcd2 = o3d.geometry.PointCloud()
        pcd2.points = o3d.utility.Vector3dVector(tgt)

        color = np.array([[0, 0, 1]] * points_numpy.shape[0])
        pcd2.colors = o3d.utility.Vector3dVector(color)

        vis = o3d.visualization.Visualizer()
        vis.create_window()
        vis.add_geometry(pcd1)
        vis.add_geometry(pcd2)

        view_control = vis.get_view_control()
        view_control.set_up([0, 1, 0])
        view_control.set_front([1, 1, 1])
        view_control.set_lookat([0, 0, 0])
        view_control.set_zoom(0.5)
        vis.run()

if __name__ == '__main__':
    it = 1
    while it <= 400000:
        if it % 10000 == 0:
            test(it)
        train(it)
        it += 1