Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,72 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os, subprocess
|
2 |
+
import uuid, tempfile
|
3 |
+
import gradio as gr
|
4 |
+
from huggingface_hub import snapshot_download
|
5 |
+
|
6 |
+
os.makedirs("pretrained", exist_ok=True)
|
7 |
+
snapshot_download(
|
8 |
+
repo_id = "jiawei011/L4GM",
|
9 |
+
local_dir = "./pretrained"
|
10 |
+
)
|
11 |
+
|
12 |
+
# Folder containing example images
|
13 |
+
examples_folder = "data_test"
|
14 |
+
|
15 |
+
# Retrieve all file paths in the folder
|
16 |
+
video_examples = [
|
17 |
+
os.path.join(examples_folder, file)
|
18 |
+
for file in os.listdir(examples_folder)
|
19 |
+
if os.path.isfile(os.path.join(examples_folder, file))
|
20 |
+
]
|
21 |
+
|
22 |
+
|
23 |
+
def generate(input_video):
|
24 |
+
|
25 |
+
#--test_path data_test/otter-on-surfboard_fg.mp4
|
26 |
+
workdir = "results"
|
27 |
+
pretrained_model = "pretrained/recon.safetensors"
|
28 |
+
num_frames = 1
|
29 |
+
test_path = input_video
|
30 |
+
|
31 |
+
try:
|
32 |
+
# Run the inference command
|
33 |
+
subprocess.run(
|
34 |
+
[
|
35 |
+
"python", "infer_3d.py", "big",
|
36 |
+
f"workspace={workdir},
|
37 |
+
f"resume={pretrained_model}",
|
38 |
+
f"num_frames={num_frames}",
|
39 |
+
f"test_path={test_path}",
|
40 |
+
],
|
41 |
+
check=True
|
42 |
+
)
|
43 |
+
|
44 |
+
|
45 |
+
# Retrieve the file name without the extension
|
46 |
+
#removed_bg_file_name = os.path.splitext(os.path.basename(removed_bg_path))[0]
|
47 |
+
output_videos = glob(os.path.join(f"{workdir}", "*.mp4"))
|
48 |
+
return output_videos
|
49 |
+
except subprocess.CalledProcessError as e:
|
50 |
+
return f"Error during inference: {str(e)}"
|
51 |
+
|
52 |
+
with gr.Blocks() as demo:
|
53 |
+
with gr.Column():
|
54 |
+
with gr.Row():
|
55 |
+
with gr.Column():
|
56 |
+
input_video = gr.Video(label="Input Video")
|
57 |
+
submit_btn = gr.Button("Submit")
|
58 |
+
with gr.Column():
|
59 |
+
output_result = gr.Video(label="Result")
|
60 |
+
|
61 |
+
gr.Examples(
|
62 |
+
examples = video_examples,
|
63 |
+
inputs = [input_video]
|
64 |
+
)
|
65 |
+
|
66 |
+
submit_btn.click(
|
67 |
+
fn = generate,
|
68 |
+
inputs = [input_video],
|
69 |
+
outputs = [output_result]
|
70 |
+
)
|
71 |
+
|
72 |
+
demo.queue().launch(show_api=False, show_error=True)
|