摘要:这节课,我翘了两天,原因是最近压力比较大。另外,这是上,下一个我们将这个有趣的项目应用到图片中,生成一些比较魔性的图片。
这节课,我翘了两天,原因是最近压力比较大。
大家可能对卷积云里雾里,这节课我们就可视化一下卷积层。通过噪声图像起点单层网络单通道/单层网络多通道/多层网络全通道 来生成几幅图像,让大家看一下卷积神经网络中某一个层的输出,通过与原图像的对比,可以看到卷积层对图像的影响。
另外,这是上,下一个我们将这个有趣的项目应用到图片中,生成一些比较魔性的图片。
选择了远方,就走下去吧!!
下一个是本项目的下一篇
必看:
本项目并不训练模型,而是“训练”图像的像素值
Deep Dream项目是Google在2015年公布的一个十分有趣的项目,在训练好的神经网络中,只需要修改几个参数就可以通过这项技术生成一张奇幻的图像。
在固定的模型上,通过不断修改图像的像素值不断改变输入,激活某一层(让其损失最大)
import tensorflow as tftf.__version__
"2.6.0"
# 检测是否支持GPUtf.test.is_gpu_available()
True
import numpy as npimport IPython.display as displayimport PIL.Imagefrom tensorflow.keras.preprocessing import image
# 图像标准化def normalize_image(img): img = 255 * (img + 1.0) / 2.0 return tf.cast(img,tf.uint8)
# 图像可视化import matplotlib.pyplot as pltdef show_image(img): display.display(PIL.Image.fromarray(np.array(img))) # plt.imshow(np.array(img))
# 保存图像文件def save_image(img,file_name): PIL.Image.fromarray(np.array(img)).save(file_name)
img_noise = np.random.uniform(size=(300,300,3)) + 100.0 # 这里加100的作用是为了提高图像的亮度img_noise = img_noise.astype(np.float32) # dtypy转换成float32show_image(normalize_image(img_noise))
base_model = tf.keras.applications.InceptionV3(include_top=False,weights="imagenet")
base_model.summary()
Model: "inception_v3"__________________________________________________________________________________________________Layer (type) Output Shape Param # Connected to ==================================================================================================input_7 (InputLayer) [(None, None, None, 0 __________________________________________________________________________________________________conv2d_564 (Conv2D) (None, None, None, 3 864 input_7[0][0] __________________________________________________________________________________________________batch_normalization_564 (BatchN (None, None, None, 3 96 conv2d_564[0][0] __________________________________________________________________________________________________activation_564 (Activation) (None, None, None, 3 0 batch_normalization_564[0][0] __________________________________________________________________________________________________conv2d_565 (Conv2D) (None, None, None, 3 9216 activation_564[0][0] __________________________________________________________________________________________________batch_normalization_565 (BatchN (None, None, None, 3 96 conv2d_565[0][0] __________________________________________________________________________________________________activation_565 (Activation) (None, None, None, 3 0 batch_normalization_565[0][0] __________________________________________________________________________________________________conv2d_566 (Conv2D) (None, None, None, 6 18432 activation_565[0][0] __________________________________________________________________________________________________batch_normalization_566 (BatchN (None, None, None, 6 192 conv2d_566[0][0] __________________________________________________________________________________________________activation_566 (Activation) (None, None, None, 6 0 batch_normalization_566[0][0] __________________________________________________________________________________________________max_pooling2d_24 (MaxPooling2D) (None, None, None, 6 0 activation_566[0][0] __________________________________________________________________________________________________conv2d_567 (Conv2D) (None, None, None, 8 5120 max_pooling2d_24[0][0] __________________________________________________________________________________________________batch_normalization_567 (BatchN (None, None, None, 8 240 conv2d_567[0][0] __________________________________________________________________________________________________activation_567 (Activation) (None, None, None, 8 0 batch_normalization_567[0][0] __________________________________________________________________________________________________conv2d_568 (Conv2D) (None, None, None, 1 138240 activation_567[0][0] __________________________________________________________________________________________________batch_normalization_568 (BatchN (None, None, None, 1 576 conv2d_568[0][0] __________________________________________________________________________________________________activation_568 (Activation) (None, None, None, 1 0 batch_normalization_568[0][0] __________________________________________________________________________________________________max_pooling2d_25 (MaxPooling2D) (None, None, None, 1 0 activation_568[0][0] __________________________________________________________________________________________________conv2d_572 (Conv2D) (None, None, None, 6 12288 max_pooling2d_25[0][0] __________________________________________________________________________________________________batch_normalization_572 (BatchN (None, None, None, 6 192 conv2d_572[0][0] __________________________________________________________________________________________________activation_572 (Activation) (None, None, None, 6 0 batch_normalization_572[0][0] __________________________________________________________________________________________________conv2d_570 (Conv2D) (None, None, None, 4 9216 max_pooling2d_25[0][0] __________________________________________________________________________________________________conv2d_573 (Conv2D) (None, None, None, 9 55296 activation_572[0][0] __________________________________________________________________________________________________batch_normalization_570 (BatchN (None, None, None, 4 144 conv2d_570[0][0] __________________________________________________________________________________________________batch_normalization_573 (BatchN (None, None, None, 9 288 conv2d_573[0][0] __________________________________________________________________________________________________activation_570 (Activation) (None, None, None, 4 0 batch_normalization_570[0][0] __________________________________________________________________________________________________activation_573 (Activation) (None, None, None, 9 0 batch_normalization_573[0][0] __________________________________________________________________________________________________average_pooling2d_54 (AveragePo (None, None, None, 1 0 max_pooling2d_25[0][0] __________________________________________________________________________________________________conv2d_569 (Conv2D) (None, None, None, 6 12288 max_pooling2d_25[0][0] __________________________________________________________________________________________________conv2d_571 (Conv2D) (None, None, None, 6 76800 activation_570[0][0] __________________________________________________________________________________________________conv2d_574 (Conv2D) (None, None, None, 9 82944 activation_573[0][0] __________________________________________________________________________________________________conv2d_575 (Conv2D) (None, None, None, 3 6144 average_pooling2d_54[0][0] __________________________________________________________________________________________________batch_normalization_569 (BatchN (None, None, None, 6 192 conv2d_569[0][0] __________________________________________________________________________________________________batch_normalization_571 (BatchN (None, None, None, 6 192 conv2d_571[0][0] __________________________________________________________________________________________________batch_normalization_574 (BatchN (None, None, None, 9 288 conv2d_574[0][0] __________________________________________________________________________________________________batch_normalization_575 (BatchN (None, None, None, 3 96 conv2d_575[0][0] __________________________________________________________________________________________________activation_569 (Activation) (None, None, None, 6 0 batch_normalization_569[0][0] __________________________________________________________________________________________________activation_571 (Activation) (None, None, None, 6 0 batch_normalization_571[0][0] __________________________________________________________________________________________________activation_574 (Activation) (None, None, None, 9 0 batch_normalization_574[0][0] __________________________________________________________________________________________________activation_575 (Activation) (None, None, None, 3 0 batch_normalization_575[0][0] __________________________________________________________________________________________________mixed0 (Concatenate) (None, None, None, 2 0 activation_569[0][0] activation_571[0][0] activation_574[0][0] activation_575[0][0] __________________________________________________________________________________________________conv2d_579 (Conv2D) (None, None, None, 6 16384 mixed0[0][0] __________________________________________________________________________________________________batch_normalization_579 (BatchN (None, None, None, 6 192 conv2d_579[0][0] __________________________________________________________________________________________________activation_579 (Activation) (None, None, None, 6 0 batch_normalization_579[0][0] __________________________________________________________________________________________________conv2d_577 (Conv2D) (None, None, None, 4 12288 mixed0[0][0] __________________________________________________________________________________________________conv2d_580 (Conv2D) (None, None, None, 9 55296 activation_579[0][0] __________________________________________________________________________________________________batch_normalization_577 (BatchN (None, None, None, 4 144 conv2d_577[0][0] __________________________________________________________________________________________________batch_normalization_580 (BatchN (None, None, None, 9 288 conv2d_580[0][0] __________________________________________________________________________________________________activation_577 (Activation) (None, None, None, 4 0 batch_normalization_577[0][0] __________________________________________________________________________________________________activation_580 (Activation) (None, None, None, 9 0 batch_normalization_580[0][0] __________________________________________________________________________________________________average_pooling2d_55 (AveragePo (None, None, None, 2 0 mixed0[0][0] __________________________________________________________________________________________________conv2d_576 (Conv2D) (None, None, None, 6 16384 mixed0[0][0] __________________________________________________________________________________________________conv2d_578 (Conv2D) (None, None, None, 6 76800 activation_577[0][0] __________________________________________________________________________________________________conv2d_581 (Conv2D) (None, None, None, 9 82944 activation_580[0][0] __________________________________________________________________________________________________conv2d_582 (Conv2D) (None, None, None, 6 16384 average_pooling2d_55[0][0] __________________________________________________________________________________________________batch_normalization_576 (BatchN (None, None, None, 6 192 conv2d_576[0][0] __________________________________________________________________________________________________batch_normalization_578 (BatchN (None, None, None, 6 192 conv2d_578[0][0] __________________________________________________________________________________________________batch_normalization_581 (BatchN (None, None, None, 9 288 conv2d_581[0][0] __________________________________________________________________________________________________batch_normalization_582 (BatchN (None, None, None, 6 192 conv2d_582[0][0] __________________________________________________________________________________________________activation_576 (Activation) (None, None, None, 6 0 batch_normalization_576[0][0] __________________________________________________________________________________________________activation_578 (Activation) (None, None, None, 6 0 batch_normalization_578[0][0] __________________________________________________________________________________________________activation_581 (Activation) (None, None, None, 9 0 batch_normalization_581[0][0] __________________________________________________________________________________________________activation_582 (Activation) (None, None, None, 6 0 batch_normalization_582[0][0] __________________________________________________________________________________________________mixed1 (Concatenate) (None, None, None, 2 0 activation_576[0][0] activation_578[0][0] activation_581[0][0] activation_582[0][0] __________________________________________________________________________________________________conv2d_586 (Conv2D) (None, None, None, 6 18432 mixed1[0][0] __________________________________________________________________________________________________batch_normalization_586 (BatchN (None, None, None, 6 192 conv2d_586[0][0] __________________________________________________________________________________________________activation_586 (Activation) (None, None, None, 6 0 batch_normalization_586[0][0] __________________________________________________________________________________________________conv2d_584 (Conv2D) (None, None, None, 4 13824 mixed1[0][0] __________________________________________________________________________________________________conv2d_587 (Conv2D) (None, None, None, 9 55296 activation_586[0][0] __________________________________________________________________________________________________batch_normalization_584 (BatchN (None, None, None, 4 144 conv2d_584[0][0] __________________________________________________________________________________________________batch_normalization_587 (BatchN (None, None, None, 9 288 conv2d_587[0][0] __________________________________________________________________________________________________activation_584 (Activation) (None, None, None, 4 0 batch_normalization_584[0][0] __________________________________________________________________________________________________activation_587 (Activation) (None, None, None, 9 0 batch_normalization_587[0][0] __________________________________________________________________________________________________average_pooling2d_56 (AveragePo (None, None, None, 2 0 mixed1[0][0] __________________________________________________________________________________________________conv2d_583 (Conv2D) (None, None, None, 6 18432 mixed1[0][0] __________________________________________________________________________________________________conv2d_585 (Conv2D) (None, None, None, 6 76800 activation_584[0][0] __________________________________________________________________________________________________conv2d_588 (Conv2D) (None, None, None, 9 82944 activation_587[0][0] __________________________________________________________________________________________________conv2d_589 (Conv2D) (None, None, None, 6 18432 average_pooling2d_56[0][0] __________________________________________________________________________________________________batch_normalization_583 (BatchN (None, None, None, 6 192 conv2d_583[0][0] __________________________________________________________________________________________________batch_normalization_585 (BatchN (None, None, None, 6 192 conv2d_585[0][0] __________________________________________________________________________________________________batch_normalization_588 (BatchN (None, None, None, 9 288 conv2d_588[0][0] __________________________________________________________________________________________________batch_normalization_589 (BatchN (None, None, None, 6 192 conv2d_589[0][0] __________________________________________________________________________________________________activation_583 (Activation) (None, None, None, 6 0 batch_normalization_583[0][0] __________________________________________________________________________________________________activation_585 (Activation) (None, None, None, 6 0 batch_normalization_585[0][0] __________________________________________________________________________________________________activation_588 (Activation) (None, None, None, 9 0 batch_normalization_588[0][0] __________________________________________________________________________________________________activation_589 (Activation) (None, None, None, 6 0 batch_normalization_589[0][0] __________________________________________________________________________________________________mixed2 (Concatenate) (None, None, None, 2 0 activation_583[0][0] activation_585[0][0] activation_588[0][0] activation_589[0][0] __________________________________________________________________________________________________conv2d_591 (Conv2D) (None, None, None, 6 18432 mixed2[0][0] __________________________________________________________________________________________________batch_normalization_591 (BatchN (None, None, None, 6 192 conv2d_591[0][0] __________________________________________________________________________________________________activation_591 (Activation) (None, None, None, 6 0 batch_normalization_591[0][0] __________________________________________________________________________________________________conv2d_592 (Conv2D) (None, None, None, 9 55296 activation_591[0][0] __________________________________________________________________________________________________batch_normalization_592 (BatchN (None, None, None, 9 288 conv2d_592[0][0] __________________________________________________________________________________________________activation_592 (Activation) (None, None, None, 9 0 batch_normalization_592[0][0] __________________________________________________________________________________________________conv2d_590 (Conv2D) (None, None, None, 3 995328 mixed2[0][0] __________________________________________________________________________________________________conv2d_593 (Conv2D) (None, None, None, 9 82944 activation_592[0][0] __________________________________________________________________________________________________batch_normalization_590 (BatchN (None, None, None, 3 1152 conv2d_590[0][0] __________________________________________________________________________________________________batch_normalization_593 (BatchN (None, None, None, 9 288 conv2d_593[0][0] __________________________________________________________________________________________________activation_590 (Activation) (None, None, None, 3 0 batch_normalization_590[0][0] __________________________________________________________________________________________________activation_593 (Activation) (None, None, None, 9 0 batch_normalization_593[0][0] __________________________________________________________________________________________________max_pooling2d_26 (MaxPooling2D) (None, None, None, 2 0 mixed2[0][0] __________________________________________________________________________________________________mixed3 (Concatenate) (None, None, None, 7 0 activation_590[0][0] activation_593[0][0] max_pooling2d_26[0][0] __________________________________________________________________________________________________conv2d_598 (Conv2D) (None, None, None, 1 98304 mixed3[0][0] __________________________________________________________________________________________________batch_normalization_598 (BatchN (None, None, None, 1 384 conv2d_598[0][0] __________________________________________________________________________________________________activation_598 (Activation) (None, None, None, 1 0 batch_normalization_598[0][0] __________________________________________________________________________________________________conv2d_599 (Conv2D) (None, None, None, 1 114688 activation_598[0][0] __________________________________________________________________________________________________batch_normalization_599 (BatchN (None, None, None, 1 384 conv2d_599[0][0] __________________________________________________________________________________________________activation_599 (Activation) (None, None, None, 1 0 batch_normalization_599[0][0] __________________________________________________________________________________________________conv2d_595 (Conv2D) (None, None, None, 1 98304 mixed3[0][0] __________________________________________________________________________________________________conv2d_600 (Conv2D) (None, None, None, 1 114688 activation_599[0][0] __________________________________________________________________________________________________batch_normalization_595 (BatchN (None, None, None, 1 384 conv2d_595[0][0] __________________________________________________________________________________________________batch_normalization_600 (BatchN (None, None, None, 1 384 conv2d_600[0][0] __________________________________________________________________________________________________activation_595 (Activation) (None, None, None, 1 0 batch_normalization_595[0][0] __________________________________________________________________________________________________activation_600 (Activation) (None, None, None, 1 0 batch_normalization_600[0][0] __________________________________________________________________________________________________conv2d_596 (Conv2D) (None, None, None, 1 114688 activation_595[0][0] __________________________________________________________________________________________________conv2d_601 (Conv2D) (None, None, None, 1 114688 activation_600[0][0] __________________________________________________________________________________________________batch_normalization_596 (BatchN (None, None, None, 1 384 conv2d_596[0][0] __________________________________________________________________________________________________batch_normalization_601 (BatchN (None, None, None, 1 384 conv2d_601[0][0] __________________________________________________________________________________________________activation_596 (Activation) (None, None, None, 1 0 batch_normalization_596[0][0] __________________________________________________________________________________________________activation_601 (Activation) (None, None, None, 1 0 batch_normalization_601[0][0] __________________________________________________________________________________________________average_pooling2d_57 (AveragePo (None, None, None, 7 0 mixed3[0][0] __________________________________________________________________________________________________conv2d_594 (Conv2D) (None, None, None, 1 147456 mixed3[0][0] __________________________________________________________________________________________________conv2d_597 (Conv2D) (None, None, None, 1 172032 activation_596[0][0] __________________________________________________________________________________________________conv2d_602 (Conv2D) (None, None, None, 1 172032 activation_601[0][0] __________________________________________________________________________________________________conv2d_603 (Conv2D) (None, None, None, 1 147456 average_pooling2d_57[0][0] __________________________________________________________________________________________________batch_normalization_594 (BatchN (None, None, None, 1 576 conv2d_594[0][0] __________________________________________________________________________________________________batch_normalization_597 (BatchN (None, None, None, 1 576 conv2d_597[0][0] __________________________________________________________________________________________________batch_normalization_602 (BatchN (None, None, None, 1 576 conv2d_602[0][0] __________________________________________________________________________________________________batch_normalization_603 (BatchN (None, None, None, 1 576 conv2d_603[0][0] __________________________________________________________________________________________________activation_594 (Activation) (None, None, None, 1 0 batch_normalization_594[0][0] __________________________________________________________________________________________________activation_597 (Activation) (None, None, None, 1 0 batch_normalization_597[0][0] __________________________________________________________________________________________________activation_602 (Activation) (None, None, None, 1 0 batch_normalization_602[0][0] __________________________________________________________________________________________________activation_603 (Activation) (None, None, None, 1 0 batch_normalization_603[0][0] __________________________________________________________________________________________________mixed4 (Concatenate) (None, None, None, 7 0 activation_594[0][0] activation_597[0][0] activation_602[0][0] activation_603[0][0] __________________________________________________________________________________________________conv2d_608 (Conv2D) (None, None, None, 1 122880 mixed4[0][0] __________________________________________________________________________________________________batch_normalization_608 (BatchN (None, None, None, 1 480 conv2d_608[0][0] __________________________________________________________________________________________________activation_608 (Activation) (None, None, None, 1 0 batch_normalization_608[0][0] __________________________________________________________________________________________________conv2d_609 (Conv2D) (None, None, None, 1 179200 activation_608[0][0] __________________________________________________________________________________________________batch_normalization_609 (BatchN (None, None, None, 1 480 conv2d_609[0][0] __________________________________________________________________________________________________activation_609 (Activation) (None, None, None, 1 0 batch_normalization_609[0][0] __________________________________________________________________________________________________conv2d_605 (Conv2D) (None, None, None, 1 122880 mixed4[0][0] __________________________________________________________________________________________________conv2d_610 (Conv2D) (None, None, None, 1 179200 activation_609[0][0] __________________________________________________________________________________________________batch_normalization_605 (BatchN (None, None, None, 1 480 conv2d_605[0][0] __________________________________________________________________________________________________batch_normalization_610 (BatchN (None, None, None, 1 480 conv2d_610[0][0] __________________________________________________________________________________________________activation_605 (Activation) (None, None, None, 1 0 batch_normalization_605[0][0] __________________________________________________________________________________________________activation_610 (Activation) (None, None, None, 1 0 batch_normalization_610[0][0] __________________________________________________________________________________________________conv2d_606 (Conv2D) (None, None, None, 1 179200 activation_605[0][0] __________________________________________________________________________________________________conv2d_611 (Conv2D) (None, None, None, 1 179200 activation_610[0][0] __________________________________________________________________________________________________batch_normalization_606 (BatchN (None, None, None, 1 480 conv2d_606[0][0] __________________________________________________________________________________________________batch_normalization_611 (BatchN (None, None, None, 1 480 conv2d_611[0][0] __________________________________________________________________________________________________activation_606 (Activation) (None, None, None, 1 0 batch_normalization_606[0][0] __________________________________________________________________________________________________activation_611 (Activation) (None, None, None, 1 0 batch_normalization_611[0][0] __________________________________________________________________________________________________average_pooling2d_58 (AveragePo (None, None, None, 7 0 mixed4[0][0] __________________________________________________________________________________________________conv2d_604 (Conv2D) (None, None, None, 1 147456 mixed4[0][0] __________________________________________________________________________________________________conv2d_607 (Conv2D) (None, None, None, 1 215040 activation_606[0][0] __________________________________________________________________________________________________conv2d_612 (Conv2D) (None, None, None, 1 215040 activation_611[0][0] __________________________________________________________________________________________________conv2d_613 (Conv2D) (None, None, None, 1 147456 average_pooling2d_58[0][0] __________________________________________________________________________________________________batch_normalization_604 (BatchN (None, None, None, 1 576 conv2d_604[0][0] __________________________________________________________________________________________________batch_normalization_607 (BatchN (None, None, None, 1 576 conv2d_607[0][0] __________________________________________________________________________________________________batch_normalization_612 (BatchN (None, None, None, 1 576 conv2d_612[0][0] __________________________________________________________________________________________________batch_normalization_613 (BatchN (None, None, None, 1 576 conv2d_613[0][0] __________________________________________________________________________________________________activation_604 (Activation) (None, None, None, 1 0 batch_normalization_604[0][0] __________________________________________________________________________________________________activation_607 (Activation) (None, None, None, 1 0 batch_normalization_607[0][0] __________________________________________________________________________________________________activation_612 (Activation) (None, None, None, 1 0 batch_normalization_612[0][0] __________________________________________________________________________________________________activation_613 (Activation) (None, None, None, 1 0 batch_normalization_613[0][0] __________________________________________________________________________________________________mixed5 (Concatenate) (None, None, None, 7 0 activation_604[0][0] activation_607[0][0] activation_612[0][0] activation_613[0][0] __________________________________________________________________________________________________conv2d_618 (Conv2D) (None, None, None, 1 122880 mixed5[0][0] __________________________________________________________________________________________________batch_normalization_618 (BatchN (None, None, None, 1 480 conv2d_618[0][0] __________________________________________________________________________________________________activation_618 (Activation) (None, None, None, 1 0 batch_normalization_618[0][0] __________________________________________________________________________________________________conv2d_619 (Conv2D) (None, None, None, 1 179200 activation_618[0][0] __________________________________________________________________________________________________batch_normalization_619 (BatchN (None, None, None, 1 480 conv2d_619[0][0] __________________________________________________________________________________________________activation_619 (Activation) (None, None, None, 1 0 batch_normalization_619[0][0] __________________________________________________________________________________________________conv2d_615 (Conv2D) (None, None, None, 1 122880 mixed5[0][0] __________________________________________________________________________________________________conv2d_620 (Conv2D) (None, None, None, 1 179200 activation_619[0][0] __________________________________________________________________________________________________batch_normalization_615 (BatchN (None, None, None, 1 480 conv2d_615[0][0] __________________________________________________________________________________________________batch_normalization_620 (BatchN (None, None, None, 1 480 conv2d_620[0][0] __________________________________________________________________________________________________activation_615 (Activation) (None, None, None, 1 0 batch_normalization_615[0][0] __________________________________________________________________________________________________activation_620 (Activation) (None, None, None, 1 0 batch_normalization_620[0][0] __________________________________________________________________________________________________conv2d_616 (Conv2D) (None, None, None, 1 179200 activation_615[0][0] __________________________________________________________________________________________________conv2d_621 (Conv2D) (None, None, None, 1 179200 activation_620[0][0] __________________________________________________________________________________________________batch_normalization_616 (BatchN (None, None, None, 1 480 conv2d_616[0][0] __________________________________________________________________________________________________batch_normalization_621 (BatchN (None, None, None, 1 480 conv2d_621[0][0] __________________________________________________________________________________________________activation_616 (Activation) (None, None, None, 1 0 batch_normalization_616[0][0] __________________________________________________________________________________________________activation_621 (Activation) (None, None, None, 1 0 batch_normalization_621[0][0] __________________________________________________________________________________________________average_pooling2d_59 (AveragePo (None, None, None, 7 0 mixed5[0][0] __________________________________________________________________________________________________conv2d_614 (Conv2D) (None, None, None, 1 147456 mixed5[0][0] __________________________________________________________________________________________________conv2d_617 (Conv2D) (None, None, None, 1 215040 activation_616[0][0] __________________________________________________________________________________________________conv2d_622 (Conv2D) (None, None, None, 1 215040 activation_621[0][0] __________________________________________________________________________________________________conv2d_623 (Conv2D) (None, None, None, 1 147456 average_pooling2d_59[0][0] __________________________________________________________________________________________________batch_normalization_614 (BatchN (None, None, None, 1 576 conv2d_614[0][0] __________________________________________________________________________________________________batch_normalization_617 (BatchN (None, None, None, 1 576 conv2d_617[0][0] __________________________________________________________________________________________________batch_normalization_622 (BatchN (None, None, None, 1 576 conv2d_622[0][0] __________________________________________________________________________________________________batch_normalization_623 (BatchN (None, None, None, 1 576 conv2d_623[0][0] __________________________________________________________________________________________________activation_614 (Activation) (None, None, None, 1 0 batch_normalization_614[0][0] __________________________________________________________________________________________________activation_617 (Activation) (None, None, None, 1 0 batch_normalization_617[0][0] __________________________________________________________________________________________________activation_622 (Activation) (None, None, None, 1 0 batch_normalization_622[0][0] __________________________________________________________________________________________________activation_623 (Activation) (None, None, None, 1 0 batch_normalization_623[0][0] __________________________________________________________________________________________________mixed6 (Concatenate) (None, None, None, 7 0 activation_614[0][0] activation_617[0][0] activation_622[0][0] activation_623[0][0] __________________________________________________________________________________________________conv2d_628 (Conv2D) (None, None, None, 1 147456 mixed6[0][0] __________________________________________________________________________________________________batch_normalization_628 (BatchN (None, None, None, 1 576 conv2d_628[0][0] __________________________________________________________________________________________________activation_628 (Activation) (None, None, None, 1 0 batch_normalization_628[0][0] __________________________________________________________________________________________________conv2d_629 (Conv2D) (None, None, None, 1 258048 activation_628[0][0] __________________________________________________________________________________________________batch_normalization_629 (BatchN (None, None, None, 1 576 conv2d_629[0][0] __________________________________________________________________________________________________activation_629 (Activation) (None, None, None, 1 0 batch_normalization_629[0][0] __________________________________________________________________________________________________conv2d_625 (Conv2D) (None, None, None, 1 147456 mixed6[0][0] __________________________________________________________________________________________________conv2d_630 (Conv2D) (None, None, None, 1 258048 activation_629[0][0] __________________________________________________________________________________________________batch_normalization_625 (BatchN (None, None, None, 1 576 conv2d_625[0][0] __________________________________________________________________________________________________batch_normalization_630 (BatchN (None, None, None, 1 576 conv2d_630[0][0] __________________________________________________________________________________________________activation_625 (Activation) (None, None, None, 1 0 batch_normalization_625[0][0] __________________________________________________________________________________________________activation_630 (Activation) (None, None, None, 1 0 batch_normalization_630[0][0] __________________________________________________________________________________________________conv2d_626 (Conv2D) (None, None, None, 1 258048 activation_625[0][0] __________________________________________________________________________________________________conv2d_631 (Conv2D) (None, None, None, 1 258048 activation_630[0][0] __________________________________________________________________________________________________batch_normalization_626 (BatchN (None, None, None, 1 576 conv2d_626[0][0] __________________________________________________________________________________________________batch_normalization_631 (BatchN (None, None, None, 1 576 conv2d_631[0][0] __________________________________________________________________________________________________activation_626 (Activation) (None, None, None, 1 0 batch_normalization_626[0][0] __________________________________________________________________________________________________activation_631 (Activation) (None, None, None, 1 0 batch_normalization_631[0][0] __________________________________________________________________________________________________average_pooling2d_60 (AveragePo (None, None, None, 7 0 mixed6[0][0] __________________________________________________________________________________________________conv2d_624 (Conv2D) (None, None, None, 1 147456 mixed6[0][0] __________________________________________________________________________________________________conv2d_627 (Conv2D) (None, None, None, 1 258048 activation_626[0][0] __________________________________________________________________________________________________conv2d_632 (Conv2D) (None, None, None, 1 258048 activation_631[0][0] __________________________________________________________________________________________________conv2d_633 (Conv2D) (None, None, None, 1 147456 average_pooling2d_60[0][0] __________________________________________________________________________________________________batch_normalization_624 (BatchN (None, None, None, 1 576 conv2d_624[0][0] __________________________________________________________________________________________________batch_normalization_627 (BatchN (None, None, None, 1 576 conv2d_627[0][0] __________________________________________________________________________________________________batch_normalization_632 (BatchN (None, None, None, 1 576 conv2d_632[0][0] __________________________________________________________________________________________________batch_normalization_633 (BatchN (None, None, None, 1 576 conv2d_633[0][0] __________________________________________________________________________________________________activation_624 (Activation) (None, None, None, 1 0 batch_normalization_624[0][0] __________________________________________________________________________________________________activation_627 (Activation) (None, None, None, 1 0 batch_normalization_627[0][0] __________________________________________________________________________________________________activation_632 (Activation) (None, None, None, 1 0 batch_normalization_632[0][0] __________________________________________________________________________________________________activation_633 (Activation) (None, None, None, 1 0 batch_normalization_633[0][0] __________________________________________________________________________________________________mixed7 (Concatenate) (None, None, None, 7 0 activation_624[0][0] activation_627[0][0] activation_632[0][0] activation_633[0][0] __________________________________________________________________________________________________conv2d_636 (Conv2D) (None, None, None, 1 147456 mixed7[0][0] __________________________________________________________________________________________________batch_normalization_636 (BatchN (None, None, None, 1 576 conv2d_636[0][0] __________________________________________________________________________________________________activation_636 (Activation) (None, None, None, 1 0 batch_normalization_636[0][0] __________________________________________________________________________________________________conv2d_637 (Conv2D) (None, None, None, 1 258048 activation_636[0][0] __________________________________________________________________________________________________batch_normalization_637 (BatchN (None, None, None, 1 576 conv2d_637[0][0] __________________________________________________________________________________________________activation_637 (Activation) (None, None, None, 1 0 batch_normalization_637[0][0] __________________________________________________________________________________________________conv2d_634 (Conv2D) (None, None, None, 1 147456 mixed7[0][0] __________________________________________________________________________________________________conv2d_638 (Conv2D) (None, None, None, 1 258048 activation_637[0][0] __________________________________________________________________________________________________batch_normalization_634 (BatchN (None, None, None, 1 576 conv2d_634[0][0] __________________________________________________________________________________________________batch_normalization_638 (BatchN (None, None, None, 1 576 conv2d_638[0][0] __________________________________________________________________________________________________activation_634 (Activation) (None, None, None, 1 0 batch_normalization_634[0][0] __________________________________________________________________________________________________activation_638 (Activation) (None, None, None, 1 0 batch_normalization_638[0][0] __________________________________________________________________________________________________conv2d_635 (Conv2D) (None, None, None, 3 552960 activation_634[0][0] __________________________________________________________________________________________________conv2d_639 (Conv2D) (None, None, None, 1 331776 activation_638[0][0] __________________________________________________________________________________________________batch_normalization_635 (BatchN (None, None, None, 3 960 conv2d_635[0][0] __________________________________________________________________________________________________batch_normalization_639 (BatchN (None, None, None, 1 576 conv2d_639[0][0] __________________________________________________________________________________________________activation_635 (Activation) (None, None, None, 3 0 batch_normalization_635[0][0] __________________________________________________________________________________________________activation_639 (Activation) (None, None, None, 1 0 batch_normalization_639[0][0] __________________________________________________________________________________________________max_pooling2d_27 (MaxPooling2D) (None, None, None, 7 0 mixed7[0][0] __________________________________________________________________________________________________mixed8 (Concatenate) (None, None, None, 1 0 activation_635[0][0] activation_639[0][0] max_pooling2d_27[0][0] __________________________________________________________________________________________________conv2d_644 (Conv2D) (None, None, None, 4 573440 mixed8[0][0] __________________________________________________________________________________________________batch_normalization_644 (BatchN (None, None, None, 4 1344 conv2d_644[0][0] __________________________________________________________________________________________________activation_644 (Activation) (None, None, None, 4 0 batch_normalization_644[0][0] __________________________________________________________________________________________________conv2d_641 (Conv2D) (None, None, None, 3 491520 mixed8[0][0] __________________________________________________________________________________________________conv2d_645 (Conv2D) (None, None, None, 3 1548288 activation_644[0][0] __________________________________________________________________________________________________batch_normalization_641 (BatchN (None, None, None, 3 1152 conv2d_641[0][0] __________________________________________________________________________________________________batch_normalization_645 (BatchN (None, None, None, 3 1152 conv2d_645[0][0] __________________________________________________________________________________________________activation_641 (Activation) (None, None, None, 3 0 batch_normalization_641[0][0] __________________________________________________________________________________________________activation_645 (Activation) (None, None, None, 3 0 batch_normalization_645[0][0] __________________________________________________________________________________________________conv2d_642 (Conv2D) (None, None, None, 3 442368 activation_641[0][0] __________________________________________________________________________________________________conv2d_643 (Conv2D) (None, None, None, 3 442368 activation_641[0][0] __________________________________________________________________________________________________conv2d_646 (Conv2D) (None, None, None, 3 442368 activation_645[0][0] __________________________________________________________________________________________________conv2d_647 (Conv2D) (None, None, None, 3 442368 activation_645[0][0] __________________________________________________________________________________________________average_pooling2d_61 (AveragePo (None, None, None, 1 0 mixed8[0][0] __________________________________________________________________________________________________conv2d_640 (Conv2D) (None, None, None, 3 409600 mixed8[0][0] __________________________________________________________________________________________________batch_normalization_642 (BatchN (None, None, None, 3 1152 conv2d_642[0][0] __________________________________________________________________________________________________batch_normalization_643 (BatchN (None, None, None, 3 1152 conv2d_643[0][0] __________________________________________________________________________________________________batch_normalization_646 (BatchN (None, None, None, 3 1152 conv2d_646[0][0] __________________________________________________________________________________________________batch_normalization_647 (BatchN (None, None, None, 3 1152 conv2d_647[0][0] __________________________________________________________________________________________________conv2d_648 (Conv2D) (None, None, None, 1 245760 average_pooling2d_61[0][0] __________________________________________________________________________________________________batch_normalization_640 (BatchN (None, None, None, 3 960 conv2d_640[0][0] __________________________________________________________________________________________________activation_642 (Activation) (None, None, None, 3 0 batch_normalization_642[0][0] __________________________________________________________________________________________________activation_643 (Activation) (None, None, None, 3 0 batch_normalization_643[0][0] __________________________________________________________________________________________________activation_646 (Activation) (None, None, None, 3 0 batch_normalization_646[0][0] __________________________________________________________________________________________________activation_647 (Activation) (None, None, None, 3 0 batch_normalization_647[0][0] __________________________________________________________________________________________________batch_normalization_648 (BatchN (None, None, None, 1 576 conv2d_648[0][0]
文章版权归作者所有,未经允许请勿转载,若此文章存在违规行为,您可以联系管理员删除。
转载请注明本文地址:https://www.ucloud.cn/yun/121807.html
摘要:机器学习进阶笔记之十那些上好玩的黑科技是基于进行研发的第二代人工智能学习系统,被广泛用于语音识别或图像识别等多项机器深度学习领域。目前的深度学习的研究领域主要有以下类人群。 特征选择 当数据预处理完成后,我们需要选择有意义的特征,输入机器学习的算法模型进行训练。 TensorFlow实现seq2seq 前言 前面在《深度学习的seq2seq模型》文章中已经介绍了seq2seq结构及其原...
摘要:诞生已经有十年之久,但是真正起势得益于去年开源了大量的深度学习模块和扩展。来自一个日本的深度学习创业公司,今年月发布的一个框架。顾名思义,是的深度学习框架,也是较早的商用级别的深度学习开源库。 本周早些时候Google开源了TensorFlow(GitHub),此举在深度学习领域影响巨大,因为Google在人工智能领域的研发成绩斐然,有着雄厚的人才储备,而且Google自己的Gmail和搜索...
摘要:学了大半年之后成功转行做前端了。包含大量其他神经网络库中的包装器和抽象,其中最值得注意的是,其中也包含一些机器学习的实用模块。它是轻量级可扩展的神经网络工具包,同时拥有友好的界面,可供机器学习的训练和预测使用。 题记:大二的时候发现人生苦短,所以信了拍神,开始学Python。学了大半年之后成功转行做前端了。来写个教程帮助大家入门Python。 Python零基础入门 零基础入门就得从最...
阅读 1339·2021-11-04 16:10
阅读 920·2021-10-08 10:05
阅读 2482·2021-09-30 09:48
阅读 2735·2019-08-29 11:31
阅读 1411·2019-08-28 18:22
阅读 3090·2019-08-26 13:44
阅读 1210·2019-08-26 13:42
阅读 2752·2019-08-26 10:20