File size: 2,475 Bytes
192d452
c7d5210
 
185a1c6
 
2c0a33a
 
 
 
 
 
 
 
 
 
 
 
054f74a
e5cf4e9
 
 
 
 
 
 
 
054f74a
185a1c6
2559312
 
5dcea5c
2559312
185a1c6
 
 
f22056c
673d632
 
 
 
185a1c6
673d632
b71e439
185a1c6
673d632
91d639b
673d632
192d452
dad1377
68dcd2b
dad1377
 
 
 
 
657b87f
 
 
 
 
 
 
 
 
dad1377
 
054f74a
4f2203a
 
 
dad1377
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
import gradio as gr
import numpy as np
from PIL import Image
from transformers import AutoImageProcessor, AutoModelForImageClassification

model_names = [
    "0-ma/swin-geometric-shapes-tiny",
    "0-ma/mobilenet-v2-geometric-shapes",
    "0-ma/focalnet-geometric-shapes-tiny",
    "0-ma/efficientnet-b2-geometric-shapes",
    "0-ma/beit-geometric-shapes-base",
    "0-ma/mit-b0-geometric-shapes",
    "0-ma/vit-geometric-shapes-base",
    "0-ma/resnet-geometric-shapes",
    "0-ma/vit-geometric-shapes-tiny",
]

example_images = [
    'example/1_None.jpg',
    'example/2_Circle.jpg',
    'example/3_Triangle.jpg',
    'example/4_Square.jpg',
    'example/5_Pentagone.jpg',
    'example/6_Hexagone.jpg'
]

labels = [example.split("_")[1].split(".")[0] for example in example_images]

feature_extractors = {model_name: AutoImageProcessor.from_pretrained(model_name) for model_name in model_names}
classification_models = {model_name: AutoModelForImageClassification.from_pretrained(model_name) for model_name in model_names}

def predict(image, selected_model):
    feature_extractor = feature_extractors[selected_model]
    model = classification_models[selected_model]

    inputs = feature_extractor(images=[image], return_tensors="pt")
    logits = model(**inputs)['logits'].cpu().detach().numpy()[0]
    logits_positive = logits
    logits_positive[logits < 0] = 0
    logits_positive = logits_positive/np.sum(logits_positive)

    confidences = {}
    for i in range(len(labels)):
        if logits[i] > 0:
            confidences[labels[i]] = float(logits_positive[i])
    return confidences

title = "Geometric Shape Classifier"
description = "Select a model and upload an image to classify geometric shapes."

with gr.Blocks() as demo:
    gr.Markdown(f"# {title}")
    gr.Markdown(description)
    
    with gr.Row():
        model_dropdown = gr.Dropdown(choices=model_names, label="Selected Model", value=model_names[0])
        gr.Examples(
            examples=example_images,
            inputs=image_input,
            outputs=output,
            fn=lambda img: predict(img, model_dropdown.value),
            cache_examples=True,
        )
    image_input = gr.Image(type="pil")
    
    output = gr.Label()

    image_input.change(fn=lambda img: predict(img, model_dropdown.value), inputs=[image_input], outputs=output)
    model_dropdown.change(fn=lambda img, model: predict(img, model), inputs=[image_input, model_dropdown], outputs=output)

demo.launch()