LuniLand/vision-classifiers

1# AUTOGENERATED! DO NOT EDIT! File to edit: app.ipynb.
2
3# %% auto 0
4__all__ = ['single_classifier', 'multi_class_classifier', 'multi_label_classifier', 'binary_labels', 'multi_class_labels',
5           'multi_label_labels', 'label_func', 'single_classification', 'multi_class_classification',
6           'multi_label_classification']
7
8# %% app.ipynb 1
9import gradio as gr
10import nbdev
11from fastai.vision.all import *
12import os
13
14# %% app.ipynb 2
15def label_func(f): return 'Cat' if f[0].isupper() else 'Dog'
16
17# %% app.ipynb 3
18single_classifier       = load_learner('models/dog-cat-classifier.pkl')
19multi_class_classifier  = load_learner('models/breeds-classifier.pkl')
20multi_label_classifier  = load_learner('models/multi-label-classification.pkl')
21
22# %% app.ipynb 4
23binary_labels = single_classifier.dls.vocab
24
25def single_classification(img):
26    img = PILImage.create(img)
27    pred, pred_idx, probs = single_classifier.predict(img)
28    return dict(zip(binary_labels, map(float, probs)))
29
30# %% app.ipynb 5
31multi_class_labels = multi_class_classifier.dls.vocab
32
33def multi_class_classification(img):
34    img = PILImage.create(img)
35    pred, pred_idx, probs = multi_class_classifier.predict(img)
36    return dict(zip(multi_class_labels, map(float, probs)))
37
38# %% app.ipynb 6
39multi_label_labels = multi_label_classifier.dls.vocab
40
41def multi_label_classification(img):
42    img = PILImage.create(img)
43    pred, pred_idx, probs = multi_label_classifier.predict(img)
44    return dict(zip(multi_label_labels, map(float, probs)))
45
46# %% app.ipynb 7
47with gr.Blocks() as demo:
48    gr.Markdown("This demo allowing you to try different vision classification models - \
49    From binary classification through multi-class and multi-label classification and finally segmentation.")
50
51    with gr.Tab("Binary"):
52        with gr.Row():
53            with gr.Column():
54                b_image_input = gr.inputs.Image(shape = (460,460))
55                with gr.Row():
56                    b_button = gr.Button("Run")
57
58                b_examples = 'models/Examples/Pets'
59                examples = gr.Examples(examples=[b_examples + '/shiba_inu_44.jpg', b_examples + '/Bengal_132.jpg',], inputs = b_image_input)
60            binary_out = gr.Label(num_top_classes=len(binary_labels))
61        
62        
63    with gr.Tab("MultiClass"):
64        with gr.Row():
65            with gr.Column():
66                m_image_input = gr.inputs.Image(shape = (460,460))
67                with gr.Row():
68                    m_button = gr.Button("Run")
69
70                m_examples = 'models/Examples/Pets'
71                examples = gr.Examples(examples=[os.path.join(m_examples, s) for s in os.listdir(m_examples) if s.endswith('jpg')], inputs = m_image_input)
72            multi_out = gr.Label(num_top_classes=len(multi_class_labels))
73        
74
75    with gr.Tab("MultiLabel"):
76        with gr.Row():
77            with gr.Column():
78                ml_image_input = gr.inputs.Image(shape = (460,460))
79                with gr.Row():
80                    ml_button = gr.Button("Run")
81                
82                ml_examples = 'models/Examples/Pascal'
83                examples = gr.Examples(examples=[os.path.join(ml_examples, s) for s in os.listdir(ml_examples) if s.endswith('jpg')], inputs = ml_image_input)
84    
85            multil_out = gr.Label(num_top_classes=len(multi_label_labels))
86        
87    b_button.click(single_classification, inputs=b_image_input, outputs=binary_out)
88    m_button.click(multi_class_classification, inputs=m_image_input, outputs=multi_out)
89    ml_button.click(multi_label_classification, inputs=ml_image_input, outputs=multil_out)
90
91demo.launch()
92