keras CNN卷積核可視化,熱度圖教程

阿新 • • 發佈：2020-06-23

卷積核可視化

import matplotlib.pyplot as plt
import numpy as np
from keras import backend as K
from keras.models import load_model

# 將浮點影象轉換成有效影象
def deprocess_image(x):
 # 對張量進行規範化
 x -= x.mean()
 x /= (x.std() + 1e-5)
 x *= 0.1
 x += 0.5
 x = np.clip(x,1)
 # 轉化到RGB陣列
 x *= 255
 x = np.clip(x,255).astype('uint8')
 return x

# 視覺化濾波器
def kernelvisual(model,layer_target=1,num_iterate=100):
 # 影象尺寸和通道
 img_height,img_width,num_channels = K.int_shape(model.input)[1:4]
 num_out = K.int_shape(model.layers[layer_target].output)[-1]

 plt.suptitle('[%s] convnet filters visualizing' % model.layers[layer_target].name)

 print('第%d層有%d個通道' % (layer_target,num_out))
 for i_kernal in range(num_out):
  input_img = model.input
  # 構建一個損耗函式，使所考慮的層的第n個濾波器的啟用最大化，-1層softmax層
  if layer_target == -1:
   loss = K.mean(model.output[:,i_kernal])
  else:
   loss = K.mean(model.layers[layer_target].output[:,:,i_kernal]) # m*28*28*128
  # 計算影象對損失函式的梯度
  grads = K.gradients(loss,input_img)[0]
  # 效用函式通過其L2範數標準化張量
  grads /= (K.sqrt(K.mean(K.square(grads))) + 1e-5)
  # 此函式返回給定輸入影象的損耗和梯度
  iterate = K.function([input_img],[loss,grads])
  # 從帶有一些隨機噪聲的灰色影象開始
  np.random.seed(0)
  # 隨機影象
  # input_img_data = np.random.randint(0,255,(1,img_height,num_channels)) # 隨機
  # input_img_data = np.zeros((1,num_channels)) # 零值
  input_img_data = np.random.random((1,num_channels)) * 20 + 128. # 隨機灰度
  input_img_data = np.array(input_img_data,dtype=float)
  failed = False
  # 執行梯度上升
  print('####################################',i_kernal + 1)
  loss_value_pre = 0
  # 執行梯度上升num_iterate步
  for i in range(num_iterate):
   loss_value,grads_value = iterate([input_img_data])
   if i % int(num_iterate/5) == 0:
    print('Iteration %d/%d,loss: %f' % (i,num_iterate,loss_value))
    print('Mean grad: %f' % np.mean(grads_value))
    if all(np.abs(grads_val) < 0.000001 for grads_val in grads_value.flatten()):
     failed = True
     print('Failed')
     break
    if loss_value_pre != 0 and loss_value_pre > loss_value:
     break
    if loss_value_pre == 0:
     loss_value_pre = loss_value
    # if loss_value > 0.99:
    #  break
   input_img_data += grads_value * 1 # e-3
  img_re = deprocess_image(input_img_data[0])
  if num_channels == 1:
   img_re = np.reshape(img_re,(img_height,img_width))
  else:
   img_re = np.reshape(img_re,num_channels))
  plt.subplot(np.ceil(np.sqrt(num_out)),np.ceil(np.sqrt(num_out)),i_kernal + 1)
  plt.imshow(img_re) #,cmap='gray'
  plt.axis('off')

 plt.show()

執行

model = load_model('train3.h5')
kernelvisual(model,-1) # 對最終輸出視覺化
kernelvisual(model,6) # 對第二個卷積層視覺化

keras CNN卷積核可視化,熱度圖教程

熱度圖

import cv2
import matplotlib.pyplot as plt
import numpy as np
from keras import backend as K
from keras.preprocessing import image

def heatmap(model,data_img,layer_idx,img_show=None,pred_idx=None):
 # 影象處理
 if data_img.shape.__len__() != 4:
  # 由於用作輸入的img需要預處理,用作顯示的img需要原圖,因此分開兩個輸入
  if img_show is None:
   img_show = data_img
  # 縮放
  input_shape = K.int_shape(model.input)[1:3]  # (28,28)
  data_img = image.img_to_array(image.array_to_img(data_img).resize(input_shape))
  # 新增一個維度->(1,224,3)
  data_img = np.expand_dims(data_img,axis=0)
 if pred_idx is None:
  # 預測
  preds = model.predict(data_img)
  # 獲取最高預測項的index
  pred_idx = np.argmax(preds[0])
 # 目標輸出估值
 target_output = model.output[:,pred_idx]
 # 目標層的輸出代表各通道關注的位置
 last_conv_layer_output = model.layers[layer_idx].output
 # 求最終輸出對目標層輸出的導數(優化目標層輸出),代表目標層輸出對結果的影響
 grads = K.gradients(target_output,last_conv_layer_output)[0]
 # 將每個通道的導數取平均,值越高代表該通道影響越大
 pooled_grads = K.mean(grads,axis=(0,1,2))
 iterate = K.function([model.input],[pooled_grads,last_conv_layer_output[0]])
 pooled_grads_value,conv_layer_output_value = iterate([data_img])
 # 將各通道關注的位置和各通道的影響乘起來
 for i in range(conv_layer_output_value.shape[-1]):
  conv_layer_output_value[:,i] *= pooled_grads_value[i]

 # 對各通道取平均得圖片位置對結果的影響
 heatmap = np.mean(conv_layer_output_value,axis=-1)
 # 規範化
 heatmap = np.maximum(heatmap,0)
 heatmap /= np.max(heatmap)
 # plt.matshow(heatmap)
 # plt.show()
 # 疊加圖片
 # 縮放成同等大小
 heatmap = cv2.resize(heatmap,(img_show.shape[1],img_show.shape[0]))
 heatmap = np.uint8(255 * heatmap)
 # 將熱圖應用於原始影象.由於opencv熱度圖為BGR,需要轉RGB
 superimposed_img = img_show + cv2.applyColorMap(heatmap,cv2.COLORMAP_JET)[:,::-1] * 0.4
 # 擷取轉uint8
 superimposed_img = np.minimum(superimposed_img,255).astype('uint8')
 return superimposed_img,heatmap
 # 顯示圖片
 # plt.imshow(superimposed_img)
 # plt.show()
 # 儲存為檔案
 # superimposed_img = img + cv2.applyColorMap(heatmap,cv2.COLORMAP_JET) * 0.4
 # cv2.imwrite('ele.png',superimposed_img)

# 生成所有卷積層的熱度圖
def heatmaps(model,img_show=None):
 if img_show is None:
  img_show = np.array(data_img)
 # Resize
 input_shape = K.int_shape(model.input)[1:3] # (28,28,1)
 data_img = image.img_to_array(image.array_to_img(data_img).resize(input_shape))
 # 新增一個維度->(1,3)
 data_img = np.expand_dims(data_img,axis=0)
 # 預測
 preds = model.predict(data_img)
 # 獲取最高預測項的index
 pred_idx = np.argmax(preds[0])
 print("預測為:%d(%f)" % (pred_idx,preds[0][pred_idx]))
 indexs = []
 for i in range(model.layers.__len__()):
  if 'conv' in model.layers[i].name:
   indexs.append(i)
 print('模型共有%d個卷積層' % indexs.__len__())
 plt.suptitle('heatmaps for each conv')
 for i in range(indexs.__len__()):
  ret = heatmap(model,indexs[i],img_show=img_show,pred_idx=pred_idx)
  plt.subplot(np.ceil(np.sqrt(indexs.__len__()*2)),np.ceil(np.sqrt(indexs.__len__()*2)),i*2 + 1)\
   .set_title(model.layers[indexs[i]].name)
  plt.imshow(ret[0])
  plt.axis('off')
  plt.subplot(np.ceil(np.sqrt(indexs.__len__()*2)),i*2 + 2)\
   .set_title(model.layers[indexs[i]].name)
  plt.imshow(ret[1])
  plt.axis('off')
 plt.show()

執行

from keras.applications.vgg16 import VGG16
from keras.applications.vgg16 import preprocess_input

model = VGG16(weights='imagenet')
data_img = image.img_to_array(image.load_img('elephant.png'))
# VGG16預處理:RGB轉BGR,並對每一個顏色通道去均值中心化
data_img = preprocess_input(data_img)
img_show = image.img_to_array(image.load_img('elephant.png'))

heatmaps(model,img_show)

elephant.png

keras CNN卷積核可視化,熱度圖教程