[视觉工程]如何可视化分析二分类模型效果

#!/usr/bin/env python
# coding: utf-8

# In[1]:


import tensorflow as tf
import numpy as np
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.applications.mobilenet import MobileNet, preprocess_input
import matplotlib
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
get_ipython().run_line_magic('matplotlib', 'inline')


# In[3]:


IMG_WIDTH, IMG_HEIGHT = 224, 224
VALIDATION_DATA_DIR = 'data/val/'
VALIDATION_BATCH_SIZE = 64

validation_datagen = ImageDataGenerator(preprocessing_function=preprocess_input)

validation_generator = validation_datagen.flow_from_directory(
    VALIDATION_DATA_DIR,
    target_size=(IMG_WIDTH, IMG_HEIGHT),
    batch_size=VALIDATION_BATCH_SIZE,
    shuffle=False,
    class_mode='categorical')


# In[4]:


from tensorflow.keras.models import load_model
from tensorflow.keras.utils import CustomObjectScope
from tensorflow.keras.initializers import glorot_uniform

with CustomObjectScope(
    {'GlorotUniform': glorot_uniform()}):
    model = load_model('data/model.h5')


# In[5]:


# Let's view the names of the files.
filenames = validation_generator.filenames
print(len(filenames))
print(filenames[:10])


# In[6]:


ground_truth = validation_generator.classes
print(ground_truth[:10])
print(len(ground_truth))


# In[7]:


label_to_index = validation_generator.class_indices
print(label_to_index)


# In[8]:


index_to_label = dict((v, k) for k, v in label_to_index.items())
print(index_to_label)


# In[9]:


predictions = model.predict_generator(validation_generator,
                                      steps=None)
print(predictions[:10])


# In[10]:


prediction_index = []
for prediction in predictions:
    prediction_index.append(np.argmax(prediction))
    
def accuracy(predictions, ground_truth):
    total = 0
    for i, j in zip(predictions, ground_truth):
        if i == j:
            total += 1
    return total * 1.0 / len(predictions)

print(accuracy(prediction_index, ground_truth))


# In[11]:


prediction_table = {}
for index, val in enumerate(predictions):
    index_of_highest_probability = np.argmax(val)
    value_of_highest_probability = val[index_of_highest_probability]
    prediction_table[index] = [
        value_of_highest_probability, index_of_highest_probability,
        ground_truth[index]
    ]
assert len(predictions) == len(ground_truth) == len(prediction_table)


# In[12]:


def get_images_with_sorted_probabilities(prediction_table,
                                         get_highest