-
Notifications
You must be signed in to change notification settings - Fork 45
/
app_svd.py
260 lines (224 loc) · 10.1 KB
/
app_svd.py
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
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
import os
import random
from argparse import ArgumentParser
import math
import gradio as gr
import torch
from diffusers.image_processor import VaeImageProcessor
from omegaconf import OmegaConf
from PIL import Image
import torchvision.transforms as T
import imageio
from diffusers import StableVideoDiffusionPipeline
from models.pipeline import TextStableVideoDiffusionPipeline
from einops import rearrange, repeat
from utils.common import read_video
css = """
.toolbutton {
margin-buttom: 0em 0em 0em 0em;
max-width: 2.5em;
min-width: 2.5em !important;
height: 2.5em;
}
"""
class AnimateController:
def __init__(self, pretrained_model_path: str, validation_data,
output_dir, motion_mask = False, motion_strength = False):
# For mixed precision training we cast the text_encoder and vae weights to half-precision
# as these models are only used for inference, keeping weights in full precision is not required.
device=torch.device("cuda")
self.validation_data = validation_data
self.output_dir = output_dir
self.pipeline = StableVideoDiffusionPipeline.from_pretrained(pretrained_model_path, torch_dtype=torch.float16, variant="fp16").to(device)
#self.pipeline = StableVideoDiffusionPipeline.from_pretrained(pretrained_model_path).to(device)
self.sample_idx = 0
def animate(
self,
init_img,
input_video,
sample_step_slider,
seed_textbox,
fps_textbox,
num_frames_textbox,
motion_bucket_id_slider,
progress=gr.Progress(),
):
if seed_textbox != "-1" and seed_textbox != "":
torch.manual_seed(int(seed_textbox))
else:
torch.seed()
seed = torch.initial_seed()
with torch.no_grad():
vae = self.pipeline.vae
validation_data = self.validation_data
validation_data.fps = int(fps_textbox)
validation_data.num_frames = int(num_frames_textbox)
validation_data.motion_bucket_id = int(motion_bucket_id_slider)
vae_processor = VaeImageProcessor()
device = vae.device
dtype = vae.dtype
f = validation_data.num_frames
pimg = Image.fromarray(init_img["background"]).convert('RGB')
np_mask = init_img["layers"][0][:,:,3]
np_mask[np_mask!=0] = 255
if np_mask.sum() == 0:
np_mask[:] = 255
if input_video is not None:
frames = read_video(input_video)
frames = [Image.fromarray(f) for f in frames]
pimg = frames[0]
width, height = pimg.size
scale = math.sqrt(width*height / (validation_data.height*validation_data.width))
block_size=64
height = round(height/scale/block_size)*block_size
width = round(width/scale/block_size)*block_size
f = len(frames)
latents = []
for frame in frames:
input_image = vae_processor.preprocess(frame, height, width)
input_image = input_image.to(dtype).to(device)
input_image_latent = vae.encode(input_image).latent_dist.mode() * vae.config.scaling_factor
latents.append(input_image_latent.unsqueeze(1))
latents = torch.cat(latents, dim=1)
else:
width, height = pimg.size
scale = math.sqrt(width*height / (validation_data.height*validation_data.width))
block_size=64
height = round(height/scale/block_size)*block_size
width = round(width/scale/block_size)*block_size
input_image = vae_processor.preprocess(pimg, height, width)
input_image = input_image.to(dtype).to(device)
input_image_latent = vae.encode(input_image).latent_dist.mode() * vae.config.scaling_factor
latents = repeat(input_image_latent, 'b c h w->b f c h w', f=f)
b, f, c, h, w = latents.shape
mask = T.ToTensor()(np_mask).to(dtype).to(device)
mask = T.Resize([h, w], antialias=False)(mask)
mask = repeat(mask, 'b h w -> b f 1 h w', f=f).detach().clone()
mask[:,0] = 0
freeze = repeat(latents[:,0], 'b c h w -> b f c h w', f=f)
condition_latents = latents * (1-mask) + freeze * mask
condition_latents = condition_latents/vae.config.scaling_factor
motion_mask = self.pipeline.unet.config.in_channels == 9
decode_chunk_size=validation_data.get("decode_chunk_size", 7)
fps=validation_data.get("fps", 7)
motion_bucket_id=validation_data.get("motion_bucket_id", 127)
if motion_mask:
video_frames = TextStableVideoDiffusionPipeline.__call__(
self.pipeline,
image=pimg,
width=width,
height=height,
num_frames=validation_data.num_frames,
num_inference_steps=validation_data.num_inference_steps,
decode_chunk_size=decode_chunk_size,
fps=fps,
motion_bucket_id=motion_bucket_id,
mask=mask,
condition_type="image",
condition_latent=condition_latents
).frames[0]
else:
video_frames = self.pipeline(
image=pimg,
width=width,
height=height,
num_frames=validation_data.num_frames,
num_inference_steps=validation_data.num_inference_steps,
fps=validation_data.fps,
decode_chunk_size=validation_data.decode_chunk_size,
motion_bucket_id=validation_data.motion_bucket_id,
).frames[0]
save_sample_path = os.path.join(
self.output_dir, f"{self.sample_idx}.mp4")
Image.fromarray(np_mask).save(os.path.join(
self.output_dir, f"{self.sample_idx}_label.jpg"))
imageio.mimwrite(save_sample_path, video_frames, fps=7)
self.sample_idx += 1
return save_sample_path
import cv2
def get_video_info(video_path):
cap = cv2.VideoCapture(video_path)
if not cap.isOpened():
return None
length = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
cap.release()
return length
def update_num_frames(input_video, num_frames_textbox):
frame_count = get_video_info(input_video)
return frame_count or 14
def ui(controller):
with gr.Blocks(css=css) as demo:
gr.HTML(
"<div align='center'><font size='7'> <img src=\"file/example/barbie2.jpg\" style=\"height: 72px;\"/ >Animate Anything For SVD</font></div>"
)
with gr.Row():
gr.Markdown(
"<div align='center'><font size='5'><a href='https://animationai.github.io/AnimateAnything'>Project Page</a>  " # noqa
"<a href='https://arxiv.org/abs/2311.12886'>Paper</a>  "
"<a href='https://github.com/alibaba/animate-anything'>Code</a>  " # noqa
)
with gr.Row(equal_height=True):
with gr.Column():
init_img = gr.ImageMask(label='Input Image', brush=gr.Brush(default_size=100))
generate_button = gr.Button(
value="Generate", variant='primary')
input_video = gr.Video(label="Input video", interactive=True)
result_video = gr.Video(
label="Generated Animation", interactive=False)
with gr.Accordion('Advance Options', open=False):
with gr.Row():
fps_textbox = gr.Number(label="Fps", value=7, minimum=1)
num_frames_textbox = gr.Number(label="Num frames", value=14, minimum=1, maximum=78)
input_video.upload(
fn=update_num_frames,
inputs=[input_video],
outputs=[num_frames_textbox]
)
motion_bucket_id_slider = gr.Slider(
label='motion_bucket_id',
value=127, step=1, minimum=0, maximum=511)
sample_step_slider = gr.Slider(
label="Sampling steps", value=25, minimum=10, maximum=100, step=1)
with gr.Row():
seed_textbox = gr.Textbox(label="Seed", value=-1)
seed_button = gr.Button(
value="\U0001F3B2", elem_classes="toolbutton")
seed_button.click(
fn=lambda x: random.randint(1, 1e8),
outputs=[seed_textbox],
queue=False
)
generate_button.click(
fn=controller.animate,
inputs=[
init_img,
input_video,
sample_step_slider,
seed_textbox,
fps_textbox,
num_frames_textbox,
motion_bucket_id_slider
],
outputs=[result_video]
)
return demo
if __name__ == "__main__":
parser = ArgumentParser()
parser.add_argument('--config', type=str, default='example/config/base.yaml')
parser.add_argument('--server-name', type=str, default='0.0.0.0')
parser.add_argument('--port', type=int, default=7860)
parser.add_argument('--share', action='store_true')
parser.add_argument('--local-debug', action='store_true')
parser.add_argument('--save-path', default='samples')
args, unknownargs = parser.parse_known_args()
LOCAL_DEBUG = args.local_debug
args_dict = OmegaConf.load(args.config)
cli_conf = OmegaConf.from_cli()
args_dict = OmegaConf.merge(args_dict, cli_conf)
controller = AnimateController(args_dict.pretrained_model_path, args_dict.validation_data,
args_dict.output_dir, args_dict.motion_mask, args_dict.motion_strength)
demo = ui(controller)
demo.queue(max_size=10)
demo.launch(server_name=args.server_name,
server_port=args.port, max_threads=40,
)