传统ZFNet可视化红外目标特征图的代码

rom keras.layers import Convolution2D, MaxPooling2D, Activation
from keras.models import Sequential
import numpy as np
import matplotlib.pyplot as plt
import cv2

tar = cv2.imread(‘D:/xlx/tar.bmp’)
plt.imshow(tar )
tar.shape

卷积层1的可视化特征图
model = Sequential()
model.add(Convolution2D(96, # number of filter layers
7, # y dimension of kernel
7, # x dimension of kernel
input_shape=tar.shape))
visualize_tar(model, tar)

池化层2的可视化特征图
model = Sequential()
model.add(MaxPooling2D(pool_size=(3,3)))
visualize_tar(model, tar)

卷积层3的可视化特征图
model = Sequential()
model.add(Convolution2D(256, # number of filter layers
5, # y dimension of kernel
5, # x dimension of kernel
input_shape=tar.shape))
visualize_tar(model, tar)

相关性太高
池化层4的可视化特征图
model = Sequential()
model.add(MaxPooling2D(pool_size=(3,3)))
visualize_tar(model, tar)

卷积层5的可视化特征图
model = Sequential()
model.add(Convolution2D(384, # number of filter layers
3, # y dimension of kernel
3, # x dimension of kernel
input_shape=tar.shape))
visualize_tar(model, tar)

图案缺乏结构性，，。
卷积层6的可视化特征图
model = Sequential()
model.add(Convolution2D(384, # number of filter layers
3, # y dimension of kernel
3, # x dimension of kernel
input_shape=tar.shape))
visualize_tar(model, tar)

卷积层7的可视化特征图
model = Sequential()
model.add(Convolution2D(256, # number of filter layers
3, # y dimension of kernel
3, # x dimension of kernel
input_shape=tar.shape))
visualize_tar(model, tar)

池化层8的可视化特征图
model = Sequential()
model.add(MaxPooling2D(pool_size=(3,3)))
visualize_tar(model, tar)

出现大量噪声字样