VGG16 (2014)| one minute summary
The original super deep ConvNet
Published in
1 min readJun 25, 2021
The 2014 paper, “Very Deep Convolutional Networks for Large-Scale Image Recognition”, from Oxford’s Visual Geometry Group (VGG) introduced what has become known as VGG16, a well known model that placed second behind Inception-v1 (GoogLeNet) at ILSVRC-2014.
- Why? Previous ConvNets (like AlexNet) had typically used pretty large convolution filters, but this limited how deep the networks could practically be.
- What? VGG16 is a typical ConvNet architecture, but one that uses a small convolution filter size and then uses the now-freed-up space to make the network really deep.
- How? VGG16 has 16 weight layers: 13 convolutional layers with 3x3 filters (the smallest size that still capture the notion of up/down, left-right, center) and some maxpool layers in between, and then 3 fully connected layers at the end.
Related articles: