Spaces:
Running
on
Zero
Running
on
Zero
File size: 3,223 Bytes
a72119e 496112d 8365126 a72119e 2189235 de54836 2189235 496112d 2189235 496112d 2189235 496112d 2189235 496112d 8365126 496112d 8365126 a72119e de54836 2189235 a72119e 496112d 2189235 496112d 2189235 496112d a72119e 496112d 2189235 496112d a72119e 2189235 a72119e 496112d 2189235 a72119e 2189235 a72119e |
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 |
import gradio as gr
from loadimg import load_img
import spaces
from transformers import AutoModelForImageSegmentation
import torch
from torchvision import transforms
import moviepy.editor as mp
from pydub import AudioSegment
from PIL import Image
import numpy as np
import os
import tempfile
import uuid
torch.set_float32_matmul_precision(["high", "highest"][0])
birefnet = AutoModelForImageSegmentation.from_pretrained(
"ZhengPeng7/BiRefNet", trust_remote_code=True
)
birefnet.to("cuda")
transform_image = transforms.Compose(
[
transforms.Resize((1024, 1024)),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
]
)
@spaces.GPU
def fn(vid, fps, color):
# Load the video using moviepy
video = mp.VideoFileClip(vid)
# Extract audio from the video
audio = video.audio
# Extract frames at the specified FPS
frames = video.iter_frames(fps=fps)
# Process each frame for background removal
processed_frames = []
for frame in frames:
pil_image = Image.fromarray(frame)
processed_image = process(pil_image, color)
processed_frames.append(np.array(processed_image))
# Create a new video from the processed frames
processed_video = mp.ImageSequenceClip(processed_frames, fps=fps)
# Add the original audio back to the processed video
processed_video = processed_video.set_audio(audio)
# Save the processed video to a temporary file
temp_dir = "temp"
os.makedirs(temp_dir, exist_ok=True)
unique_filename = str(uuid.uuid4()) + ".mp4"
temp_filepath = os.path.join(temp_dir, unique_filename)
processed_video.write_videofile(temp_filepath, codec="libx264")
# Return the path to the temporary file
return temp_filepath
def process(image, color_hex):
image_size = image.size
input_images = transform_image(image).unsqueeze(0).to("cuda")
# Prediction
with torch.no_grad():
preds = birefnet(input_images)[-1].sigmoid().cpu()
pred = preds[0].squeeze()
pred_pil = transforms.ToPILImage()(pred)
mask = pred_pil.resize(image_size)
# Convert hex color to RGB tuple
color_rgb = tuple(int(color_hex[i : i + 2], 16) for i in (1, 3, 5))
# Create a background image with the chosen color
background = Image.new("RGBA", image_size, color_rgb + (255,))
# Composite the image onto the background using the mask
image = Image.composite(image, background, mask)
return image
def process_file(f, color="#00FF00"):
name_path = f.rsplit(".", 1)[0] + ".png"
im = load_img(f, output_type="pil")
im = im.convert("RGB")
transparent = process(im, color)
transparent.save(name_path)
return name_path
with gr.Blocks() as demo:
in_video = gr.Video(label="birefnet")
out_video = gr.Video()
fps_slider = gr.Slider(minimum=1, maximum=60, step=1, value=12, label="Output FPS")
color_picker = gr.ColorPicker(label="Background Color", value="#00FF00")
submit_button = gr.Button("Process Video")
submit_button.click(
fn, inputs=[in_video, fps_slider, color_picker], outputs=out_video
)
if __name__ == "__main__":
demo.launch(show_error=True) |