TensorFlow框架案例实战-第3课-卷积神经网络与图像应用


第3课 卷积神经网络与图像应用

Image Classification Pipeline

> Image Classification: A core task in Computer Vision

Typical pipeline:

Data-driven approach!
  • Data: Collect a dataset, images and labels.
  • Training: Use neural networks to train a classifier.
  • Test: Evaluate the classifier on unseen images.


LeNet-5

LeNet-5

convolution

convolution

max pooling

max pooling

每次卷积之后,要使用一次非线性变换。

CNN Architectures

  • AlexNet (2012 Imagenet Champion)
  • VGG (2014 Imagenet Champion)
  • GoogLeNet (2014 Imagenet Champion)
  • ResNet (2015 Imagenet Champion)


model

vgg

卷积:检测 kernel 和图像之间的相似性。

GoogleNet

GoogleNet

ResNet

ResNet

跳跃,弥补梯度消散。
已邀请:

要回复问题请先登录注册

返回顶部