A guide to convolution arithmetic for deep learning

Code

• github.com/vdumoulin/conv_arithmetic
  OfficialIn papernone★ 0
• github.com/mrdbourke/tensorflow-deep-learning
  tf★ 5,869
• github.com/marbleton/FPGA_MNIST
  pytorch★ 10
• github.com/ashantanu/DCGAN
  pytorch★ 0
• github.com/atriumlts/subpixel
  tf★ 0
• github.com/smosanu/acomMNIST
  tf★ 0
• github.com/JakobKallestad/InceptionV3-on-plankton-images
  none★ 0
• github.com/ruiponte1990/svhn
  tf★ 0
• github.com/thisiskhan/tensorflow-developer-certificate-machine-learning-kit
  tf★ 0
• github.com/avillemin/GANs
  pytorch★ 0

Abstract

We introduce a guide to help deep learning practitioners understand and manipulate convolutional neural network architectures. The guide clarifies the relationship between various properties (input shape, kernel shape, zero padding, strides and output shape) of convolutional, pooling and transposed convolutional layers, as well as the relationship between convolutional and transposed convolutional layers. Relationships are derived for various cases, and are illustrated in order to make them intuitive.

Tasks

Reproductions