File size: 3,485 Bytes
64d14f7
 
 
dad1fea
 
 
64d14f7
 
0c48d27
dad1fea
0c48d27
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
dad1fea
0c48d27
 
 
 
 
 
 
dad1fea
 
0c48d27
 
 
 
 
dad1fea
0c48d27
 
dad1fea
0c48d27
 
 
 
 
 
 
 
dad1fea
0c48d27
 
 
 
 
 
 
 
64d14f7
 
0c48d27
64d14f7
 
 
 
 
 
 
23e9ea3
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
# Import the libraries
import numpy as np
import pandas as pd
from tensorflow.keras.models import load_model # type: ignore
from tensorflow.keras.preprocessing.image import load_img, img_to_array # type: ignore
from tensorflow.keras.applications.convnext import preprocess_input # type: ignore
import gradio as gr

# Load the model
model = load_model('models/ConvNeXtBase_80_tresh_spp.tf', call_endpoint='serving_default')

# Load the taxonomy .csv
taxo_df = pd.read_csv('taxonomy/taxonomy_mapping.csv', sep=';')
taxo_df['species'] = taxo_df['species'].str.replace('_', ' ')

# Extract unique class names from the 'species' column
class_names = sorted(taxo_df['species'].unique())

# Function to map predicted class index to class name
def get_class_name(predicted_class):
    return class_names[predicted_class]

# Function to load and preprocess the image
def load_and_preprocess_image(image, target_size=(224, 224)):
    # Resize the image (assuming image is a PIL image)
    img_array = img_to_array(image.resize(target_size))
    # Expand the dimensions of the array to match model input
    img_array = np.expand_dims(img_array, axis=0)
    # Preprocess using the appropriate function (for example, ResNet50)
    img_array = preprocess_input(img_array)
    return img_array

# Function to make predictions
def make_prediction(image):
    # Preprocess the image
    img_array = load_and_preprocess_image(image)
    # Make a prediction
    prediction = model.predict(img_array)
    
    # Get the top 5 predictions
    top_indices = np.argsort(prediction[0])[-5:][::-1]  # Get indices of top 5 classes
    
    # Get predicted class and common name for the top prediction
    predicted_class_index = np.argmax(prediction)
    predicted_class_name = get_class_name(predicted_class_index)
    predicted_common_name = taxo_df[taxo_df['species'] == predicted_class_name]['common_name'].values[0]  # Get common name
    confidence = prediction[0][predicted_class_index] * 100  # Confidence of the predicted class
    
    # Create output text with HTML formatting
    output_text = f"<h1 style='font-weight: bold;'><span style='font-style: italic;'>{predicted_class_name}</span> ({predicted_common_name})</h1>"  # Large bold for predicted class, italic for class name
    output_text += "<h4 style='font-weight: bold; font-size: 1.2em;'>Top 5 Predictions:</h4>"  # Bold and larger font for predictions
    
    for i in top_indices:
        class_name = get_class_name(i)
        common_name = taxo_df[taxo_df['species'] == class_name]['common_name'].values[0]  # Get common name from CSV
        confidence_percentage = prediction[0][i] * 100
        
        # Format the output with space between class name and common name
        output_text += f"<div style='display: flex; justify-content: space-between;'>" \
                       f"<span style='font-style: italic;'>{class_name}</span>&nbsp;(<span>{common_name}</span>)" \
                       f"<span style='margin-left: auto;'>{confidence_percentage:.2f}%</span></div>"

    return output_text

# Define the Gradio interface
interface = gr.Interface(
    fn=make_prediction,           # Function to be called for predictions
    inputs=gr.Image(type="pil"),  # Input type: Image (PIL format)
    outputs="html",               # Output type: HTML for formatting
    title="Amazon arboreal species classification",
    description="Upload an image to classify the species."
)

# Launch the Gradio interface
interface.launch()