Interactive Visual by deeplizard

Interactive Transposed Convolution Demo

Transposed convolutions are a particular type of upsampling technique often used in neural networks, particularly GANs. Just like other upsampling techniques, transposed convolutions increase the size of data. For image data, transposed convolutions are used to increase image dimensions. Transposed convolutions make use of filters with a defined filter size, stride, and padding. This demo assumes no padding and a stride of 1.

To learn more about transposed convolutions in neural networks, see this lesson.