Spaces:
Paused
Paused
Update app.py
Browse files
app.py
CHANGED
@@ -1,13 +1,19 @@
|
|
1 |
-
|
2 |
import gradio as gr
|
3 |
import os
|
4 |
import torch
|
5 |
from PIL import Image
|
6 |
from inference.flovd_demo import generate_video
|
7 |
-
import requests
|
8 |
-
import shutil
|
9 |
from huggingface_hub import snapshot_download
|
10 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
11 |
# Get the HF token from environment variable
|
12 |
hf_token = os.getenv("HF_TOKEN")
|
13 |
|
@@ -17,11 +23,12 @@ snapshot_download(
|
|
17 |
repo_type="dataset",
|
18 |
local_dir="./ckpt",
|
19 |
allow_patterns="ckpt/**",
|
|
|
20 |
token=hf_token
|
21 |
)
|
22 |
|
23 |
-
#
|
24 |
-
print("📂 Contents of ./ckpt directory:")
|
25 |
for root, dirs, files in os.walk("./ckpt"):
|
26 |
level = root.replace("./ckpt", "").count(os.sep)
|
27 |
indent = " " * 4 * level
|
@@ -32,7 +39,6 @@ for root, dirs, files in os.walk("./ckpt"):
|
|
32 |
|
33 |
# --------- UI Function ---------
|
34 |
def run_flovd(prompt, image, cam_pose_name):
|
35 |
-
download_if_missing()
|
36 |
image_path = "./temp_input.png"
|
37 |
image.save(image_path)
|
38 |
|
@@ -42,10 +48,16 @@ def run_flovd(prompt, image, cam_pose_name):
|
|
42 |
omsm_path=OMSM_PATH,
|
43 |
image_path=image_path,
|
44 |
cam_pose_name=cam_pose_name,
|
45 |
-
output_path=
|
|
|
|
|
|
|
|
|
46 |
dtype=torch.float16,
|
|
|
47 |
)
|
48 |
-
|
|
|
49 |
|
50 |
# --------- Launch Gradio ---------
|
51 |
iface = gr.Interface(
|
@@ -58,4 +70,5 @@ iface = gr.Interface(
|
|
58 |
outputs=gr.Video(label="Generated Video"),
|
59 |
title="FloVD - Optical Flow Video Diffusion with Camera Motion",
|
60 |
)
|
|
|
61 |
iface.launch(server_name="0.0.0.0", server_port=7860)
|
|
|
|
|
1 |
import gradio as gr
|
2 |
import os
|
3 |
import torch
|
4 |
from PIL import Image
|
5 |
from inference.flovd_demo import generate_video
|
|
|
|
|
6 |
from huggingface_hub import snapshot_download
|
7 |
|
8 |
+
# Constants
|
9 |
+
FVSM_PATH = "./ckpt/FVSM/FloVD_FVSM_Controlnet.pt"
|
10 |
+
OMSM_PATH = "./ckpt/OMSM/"
|
11 |
+
DEPTH_CKPT_PATH = "./ckpt/others/depth_anything_v2_metric_hypersim_vitb.pth"
|
12 |
+
OUTPUT_PATH = "./results/"
|
13 |
+
POSE_TYPE = "re10k"
|
14 |
+
CONTROLNET_GUIDANCE_END = 0.4
|
15 |
+
SPEED = 1.0
|
16 |
+
|
17 |
# Get the HF token from environment variable
|
18 |
hf_token = os.getenv("HF_TOKEN")
|
19 |
|
|
|
23 |
repo_type="dataset",
|
24 |
local_dir="./ckpt",
|
25 |
allow_patterns="ckpt/**",
|
26 |
+
cache_dir="./hf_cache", # avoid permission issues
|
27 |
token=hf_token
|
28 |
)
|
29 |
|
30 |
+
# Print the directory structure of ckpt
|
31 |
+
print("\n📂 Contents of ./ckpt directory:")
|
32 |
for root, dirs, files in os.walk("./ckpt"):
|
33 |
level = root.replace("./ckpt", "").count(os.sep)
|
34 |
indent = " " * 4 * level
|
|
|
39 |
|
40 |
# --------- UI Function ---------
|
41 |
def run_flovd(prompt, image, cam_pose_name):
|
|
|
42 |
image_path = "./temp_input.png"
|
43 |
image.save(image_path)
|
44 |
|
|
|
48 |
omsm_path=OMSM_PATH,
|
49 |
image_path=image_path,
|
50 |
cam_pose_name=cam_pose_name,
|
51 |
+
output_path=OUTPUT_PATH,
|
52 |
+
controlnet_guidance_end=CONTROLNET_GUIDANCE_END,
|
53 |
+
pose_type=POSE_TYPE,
|
54 |
+
speed=SPEED,
|
55 |
+
depth_ckpt_path=DEPTH_CKPT_PATH,
|
56 |
dtype=torch.float16,
|
57 |
+
use_flow_integration=True,
|
58 |
)
|
59 |
+
|
60 |
+
return "./results/generated_videos/your_video.mp4"
|
61 |
|
62 |
# --------- Launch Gradio ---------
|
63 |
iface = gr.Interface(
|
|
|
70 |
outputs=gr.Video(label="Generated Video"),
|
71 |
title="FloVD - Optical Flow Video Diffusion with Camera Motion",
|
72 |
)
|
73 |
+
|
74 |
iface.launch(server_name="0.0.0.0", server_port=7860)
|