nn pic model preprocess note

来源：互联网发布：自己如何做淘宝店铺编辑：程序博客网时间：2024/06/10 20:51

inception 预处理

https://github.com/fchollet/keras/blob/master/keras/applications/inception_v3.py

def preprocess_input(x):    x /= 255.    x -= 0.5    x *= 2.    return x

vgg 预处理

    if len(x.shape) == 3:        #print("[INFO] channel:{}".format(data_format))        if data_format == 'channels_first':            x = x.transpose(2, 0, 1)    elif len(x.shape) == 2:        if data_format == 'channels_first':            x = x.reshape((1, x.shape[0], x.shape[1]))        else:            x = x.reshape((x.shape[0], x.shape[1], 1))    else:        raise ValueError('Unsupported image shape: ', x.shape)

im[:,:,0] -= 103.939im[:,:,1] -= 116.779im[:,:,2] -= 123.68

top2

这里写图片描述

try

_________________________________________________________________Layer (type)                 Output Shape              Param #=================================================================input_1 (InputLayer)         (None, 224, 224, 3)       0_________________________________________________________________block1_conv1 (Conv2D)        (None, 224, 224, 64)      1792_________________________________________________________________block1_conv2 (Conv2D)        (None, 224, 224, 64)      36928_________________________________________________________________block1_pool (MaxPooling2D)   (None, 112, 112, 64)      0_________________________________________________________________block2_conv1 (Conv2D)        (None, 112, 112, 128)     73856_________________________________________________________________block2_conv2 (Conv2D)        (None, 112, 112, 128)     147584_________________________________________________________________block2_pool (MaxPooling2D)   (None, 56, 56, 128)       0_________________________________________________________________block3_conv1 (Conv2D)        (None, 56, 56, 256)       295168_________________________________________________________________block3_conv2 (Conv2D)        (None, 56, 56, 256)       590080_________________________________________________________________block3_conv3 (Conv2D)        (None, 56, 56, 256)       590080_________________________________________________________________block3_pool (MaxPooling2D)   (None, 28, 28, 256)       0_________________________________________________________________block4_conv1 (Conv2D)        (None, 28, 28, 512)       1180160_________________________________________________________________block4_conv2 (Conv2D)        (None, 28, 28, 512)       2359808_________________________________________________________________block4_conv3 (Conv2D)        (None, 28, 28, 512)       2359808_________________________________________________________________block4_pool (MaxPooling2D)   (None, 14, 14, 512)       0_________________________________________________________________block5_conv1 (Conv2D)        (None, 14, 14, 512)       2359808_________________________________________________________________block5_conv2 (Conv2D)        (None, 14, 14, 512)       2359808_________________________________________________________________block5_conv3 (Conv2D)        (None, 14, 14, 512)       2359808_________________________________________________________________block5_pool (MaxPooling2D)   (None, 7, 7, 512)         0_________________________________________________________________flatten (Flatten)            (None, 25088)             0_________________________________________________________________fc1 (Dense)                  (None, 4096)              102764544_________________________________________________________________fc2 (Dense)                  (None, 4096)              16781312_________________________________________________________________logit (Dense)                (None, 1)                 4097=================================================================

0 0