EasyDeepLearn

What is the design principle behind VGG?

easy

Answer

  • Stack 3x3 conv blocks repeatedly (2, 3, or 4 in a row) and downsample with max-pooling.
  • Two stacked 3x3 convs have the same receptive field as one 5x5 but with fewer parameters and one extra nonlinearity — makes deeper nets with more nonlinearity cheap.
  • VGG-16 / VGG-19 became the standard feature extractor for years.
  • Downsides: enormous FC layers (140M params in VGG-16).
Check yourself — multiple choice
  • Big 11x11 filters
  • Stack many 3x3 convs — same receptive field as bigger kernels, fewer params, more nonlinearity
  • No convolutions
  • Only 1 layer deep

VGG: 3x3 conv stacks — deep, few kernel choices, big FC head.

#architectures#computer-vision

Practise Deep Learning

214 interview questions in this topic.

Related questions