keras 热力图

主要用于对神经网络训练的参考

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
"""
keras神经网络的可视化
"""

from tensorflow.keras.applications.vgg16 import preprocess_input
from tensorflow.keras.models import load_model

from tensorflow.keras import preprocessing
from tensorflow.keras import backend as K
from tensorflow.keras import models

import tensorflow as tf
import numpy as np
import cv2
image_size = 224

model_path = "模型地址"
img_path = "图片地址"

# 加载之前训练好的模型
model = load_model(model_path)
image = cv2.imread(img_path)
img_tensor = preprocessing.image.img_to_array(image)
img_tensor = np.expand_dims(img_tensor, axis=0)
img_tensor = preprocess_input(img_tensor)

# 获取最后一层卷积，以(Conv2d)标注的最后一个
conv_layer = model.get_layer("conv2d_93")
heatmap_model = models.Model([model.inputs], [conv_layer.output, model.output])

# 获取最后一层卷积的梯度
with tf.GradientTape() as gtape:
    conv_output, predictions = heatmap_model(img_tensor)
    loss = predictions[:, np.argmax(predictions[0])]
    grads = gtape.gradient(loss, conv_output)[0]
    pooled_grads = K.mean(grads, axis=(0, 1, 2))

# 热力图归一化
heatmap = tf.reduce_mean(tf.multiply(pooled_grads, conv_output), axis=-1)
heatmap = np.maximum(heatmap, 0)
max_heat = np.max(heatmap)
if max_heat == 0:
    max_heat = 1e-10
heatmap /= max_heat
heatmap = np.uint8(255 * heatmap)


# 热力图和原图合并
img = cv2.imread(img_path)
heatmap1 = cv2.resize(heatmap[0], (img.shape[1], img.shape[0]), interpolation=cv2.INTER_CUBIC)
heatmap1 = np.uint8(255*heatmap1)
heatmap1 = cv2.applyColorMap(heatmap1, cv2.COLORMAP_JET)
frame_out = cv2.addWeighted(img, 0.5, heatmap1, 0.5, 0)
cv2.imwrite('热力图保存路径', frame_out)

print("finish")