File size: 6,976 Bytes
8fdbedb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
import matplotlib.pyplot as plt
import seaborn as sns
plt.rcParams["figure.figsize"] = (15, 10)
plt.rcParams["figure.dpi"] = 125
plt.rcParams["font.size"] = 14
plt.rcParams['font.family'] = ['sans-serif']
plt.rcParams['font.sans-serif'] = ['DejaVu Sans']
plt.style.use('ggplot')
sns.set_style("whitegrid", {'axes.grid': False})
plt.rcParams['image.cmap'] = 'gray' # grayscale looks better
from pathlib import Path
import numpy as np
import pandas as pd
import os
from skimage.io import imread as imread
from skimage.util import montage
from PIL import Image
montage_rgb = lambda x: np.stack([montage(x[:, :, :, i]) for i in range(x.shape[3])], -1)
from skimage.color import label2rgb

image_dir = Path('D:/vscodeim/food classification/feature_food/')


mapping_file = Path('D:/vscodeim/food classification/feature_food/clean_list.json')
alleg_df = pd.read_json(mapping_file)

alleg_df['image_path'] = alleg_df['image_path'].map(lambda x: image_dir / 'subset' / x) 
print(alleg_df['image_path'].map(lambda x: x.exists()).value_counts())
allergens = alleg_df.columns[3:].tolist()
alleg_df.sample(2)

import os
from pathlib import Path


color_file = Path('D:/vscodeim/food classification/feature_food/color_features.json')
color_feat_df = pd.read_json(color_file)
color_feat_df['image_path'] = color_feat_df['image_path'].map(lambda x: image_dir / 'subset' / x) 

color_feat_dict = {c_row['image_path']: c_row['color_features'] for _, c_row in color_feat_df.iterrows()}
# add a new color feature column
alleg_df['color_features'] = alleg_df['image_path'].map(color_feat_dict.get)
alleg_df.sample(2)


co_all = np.corrcoef(np.stack(alleg_df[allergens].applymap(lambda x: 1 if x>0 else 0).values, 0).T)
fig, ax1 = plt.subplots(1, 1, figsize=(10, 10))
sns.heatmap(co_all, annot=True, fmt='2.1%', ax=ax1, cmap='RdBu', vmin=-1, vmax=1)
ax1.set_xticklabels(allergens, rotation=90)
ax1.set_yticklabels(allergens);


# package the allergens together
alleg_df['allergy_vec'] = alleg_df[allergens].applymap(lambda x: 1 if x>0 else 0).values.tolist()

from sklearn.model_selection import train_test_split
train_df, valid_df = train_test_split(alleg_df.drop(columns='ingredients_list'), 
                                      test_size=0.1, 
                                      random_state=2019, 
                                      stratify=alleg_df['allergy_vec'].map(lambda x: x[0:3]))

train_df.reset_index(inplace=True)
valid_df.reset_index(inplace=True)


print(train_df.shape[0], 'training images')
print(valid_df.shape[0], 'validation images')

train_x_vec = np.stack(train_df['color_features'].values, 0)
train_y_vec = np.stack(train_df['allergy_vec'], 0)





from sklearn.pipeline import make_pipeline
from sklearn.ensemble import RandomForestRegressor
from sklearn.preprocessing import RobustScaler  # 导入 RobustScaler

rf_pipe = make_pipeline(RobustScaler(), RandomForestRegressor(n_estimators=15))
rf_pipe.fit(train_x_vec, train_y_vec)


import streamlit as st
import pandas as pd
import numpy as np
from sklearn.metrics import roc_auc_score, roc_curve, accuracy_score
import matplotlib.pyplot as plt
from imageio import imread

@st.cache
def show_model_results(in_model, picture_number=None):
    # Your data loading code here
    # Ensure 'valid_df', 'valid_x_vec', 'valid_y_vec', 'train_df', 'train_x_vec', 'train_y_vec', 'allergens' are defined

    
    
    valid_x_vec = np.stack(valid_df['color_features'].values, 0)
    valid_y_vec = np.stack(valid_df['allergy_vec'], 0)

    x_vec = valid_x_vec
    y_vec = valid_y_vec

    
    valid_pred = in_model.predict(x_vec)
    valid_num = picture_number

    fig, m_axs = plt.subplots(2, 2, figsize=(10, 10))
    all_rows = []
    ax1 = m_axs[0, 0]
    for i, c_allergen in enumerate(allergens):
        tpr, fpr, _ = roc_curve(y_vec[:, i], valid_pred[:, i])
        auc = roc_auc_score(y_vec[:, i], valid_pred[:, i])
        acc = accuracy_score(y_vec[:, i], valid_pred[:, i] > 0.5)
        ax1.plot(tpr, fpr, '.-', label='{}: AUC {:0.2f}, Accuracy: {:2.0%}'.format(c_allergen, auc, acc))
        all_rows += [{'allegen': c_allergen, 
                      'prediction': valid_pred[j, i], 
                      'class': 'Positive' if y_vec[j, i] > 0.5 else 'Negative'} 
                     for j in range(valid_pred.shape[0])]
    
    d_ax = m_axs[0, 1]
    t_yp = np.mean(valid_pred, 0)
    t_y = np.mean(y_vec, 0)
    d_ax.barh(np.arange(len(allergens)) + 0.1, t_yp, alpha=0.5, label='Predicted')
    d_ax.barh(np.arange(len(allergens)) - 0.1, t_y + 0.001, alpha=0.5, label='Ground Truth')
    d_ax.set_xlim(0, 1)
    d_ax.set_yticks(range(len(allergens)))
    d_ax.set_yticklabels(allergens, rotation=0)
    d_ax.set_title('Overall')
    d_ax.legend()
    
    ax1.legend()
    for (_, c_row), (c_ax, d_ax) in zip(
    valid_df.iloc[valid_num:valid_num+1].iterrows(), 
    m_axs[1:]):
        c_ax.imshow(imread(c_row['image_path']))
        c_ax.set_title(c_row['title'])
        c_ax.axis('off')
        t_yp = in_model.predict(np.expand_dims(c_row['color_features'], 0))
        t_y = np.array(c_row['allergy_vec'])
        d_ax.barh(np.arange(len(allergens)) + 0.1, t_yp[0], alpha=0.5, label='Predicted')
        d_ax.barh(np.arange(len(allergens)) - 0.1, t_y + 0.001, alpha=0.5, label='Ground Truth')
        d_ax.set_yticks(range(len(allergens)))
        d_ax.set_yticklabels(allergens, rotation=0)
        d_ax.set_xlim(0, 1)
        d_ax.legend()

        # 将当前图像添加到 Streamlit 页面
    st.pyplot(fig)
    return st.write("Completed")

# Assuming you have already defined 'rf_pipe' and 'valid_df' with image paths
image_paths = valid_df['image_path'].tolist()


# Streamlit app
# Streamlit app
def main():
    st.title('Model Results')
    st.write(f'<span style="font-size:20px;">This is a prototype, so we use the images from the test set as examples.</span>', unsafe_allow_html=True)

    image_paths = valid_df['image_path'].tolist()

    num_rows = 2
    num_cols = 5
    fig, axes = plt.subplots(num_rows, num_cols, figsize=(15, 6))

    for i in range(num_rows):
        for j in range(num_cols):
            index = i * num_cols + j + 21  # Starting from index 21
            image = imread(image_paths[index])
            axes[i, j].imshow(image)
            axes[i, j].axis('off')
            axes[i, j].set_title(f'Image {index-20}')

    st.pyplot(fig)

    num_images = 10 

    # User interaction to select image
    st.write(f'<span style="font-size:20px;">Enter the image number you want to analyze.</span>', unsafe_allow_html=True)

    choice = st.number_input(f"Range (1-{num_images}): ", min_value=1, max_value=num_images)
    
    # Show model results
    if st.button('Show Results'):
        show_model_results(rf_pipe, choice+20)

if __name__ == "__main__":
    main()