File size: 2,168 Bytes
cfb3fab
 
bf12e57
 
 
cfb3fab
 
bf12e57
8397178
cfb3fab
bf12e57
cfb3fab
 
 
bf12e57
a56d89f
bf12e57
a56d89f
 
bf12e57
a56d89f
 
bf12e57
a56d89f
8397178
 
 
bf12e57
 
 
 
 
 
 
 
 
 
 
 
 
cfb3fab
bf12e57
 
 
 
 
 
 
 
 
 
 
 
cfb3fab
 
 
db457c9
cfb3fab
bf12e57
 
 
 
 
 
 
 
 
cfb3fab
bf12e57
cfb3fab
 
54fea3c
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
import torch
import torch.nn.functional as F
from transformers import AutoModel, AutoImageProcessor
from PIL import Image
from rembg import remove
import gradio as gr
import spaces
import io
import numpy as np

# Load the Nomic embed model
processor = AutoImageProcessor.from_pretrained("nomic-ai/nomic-embed-vision-v1.5")
vision_model = AutoModel.from_pretrained("nomic-ai/nomic-embed-vision-v1.5", trust_remote_code=True)

def focus_on_subject(image: Image.Image) -> Image.Image:
    """
    Remove background and crop to the main object using rembg.

    Args:
        image (PIL.Image.Image): Input image.

    Returns:
        PIL.Image.Image: Cropped image with background removed.
    """
    if isinstance(image, np.ndarray):
        image = Image.fromarray(image)
        
    image = image.convert("RGB")
    
    # Remove background
    img_bytes = io.BytesIO()
    image.save(img_bytes, format="PNG")
    img_bytes = img_bytes.getvalue()
    result_bytes = remove(img_bytes)
    
    result_image = Image.open(io.BytesIO(result_bytes)).convert("RGBA")
    bbox = result_image.getbbox()
    cropped = result_image.crop(bbox) if bbox else result_image
    
    return cropped.convert("RGB")

def ImgEmbed(image: Image.Image):
    """
    Preprocess image, generate normalized embedding, and return both embedding and processed image.

    Args:
        image (PIL.Image.Image): Input image.

    Returns:
        Tuple: (embedding list, processed image)
    """
    focused_image = focus_on_subject(image)
    inputs = processor(focused_image, return_tensors="pt")
    img_emb = vision_model(**inputs).last_hidden_state
    img_embeddings = F.normalize(img_emb[:, 0], p=2, dim=1)

    return img_embeddings[0].tolist()

# Gradio UI
with gr.Blocks() as demo:
    with gr.Row():
        with gr.Column():
            img = gr.Image(label="Upload Image")
            btn = gr.Button("Get Embeddings")
        with gr.Column():
            pre_img = gr.Image(label="Preprocessed Image")
            out = gr.Text(label="Image Embedding")

    btn.click(ImgEmbed, inputs=[img], outputs=[out, pre_img])

if __name__ == "__main__":
    demo.launch(mcp_server=True)