MobilenetV2 :

It is Neural Network Architecture Which is specially Optimized for the Edge devices
like Mobile.The architecture delivers high accuracy results while keeping the
parameters and mathematical operations as low as possible to bring deep neural
networks to mobile devices.

Earlier their was older version called MobilenetV1 which was designed for the same
purpose of edge device computing.

The main goal of these two Architecture was to Reduce two factors listed below

1. Smaller model size: Fewer number of parameters

2. Smaller complexity: Fewer Multiplications and Additions (Multi-Adds)

Architecture of MobilenetV1 :

The main difference of this architecture compared to Basic Convolution was The
Concept of Depthwise Separable Convolution.
It consisted two Sub Blocks

1.The first layer is called a depthwise convolution, it performs lightweight filtering by

applying a single convolutional filter per input channel.

2. The second layer is a 1×1 convolution, called a pointwise convolution, which is

responsible for building new features through computing linear combinations of the
input channels.

And After each convolution Batch Normalization and a non linear Activation
function Relu was applied.
However MobileNetV2 an improved version of Mobilenetv1 is released which is
much faster in comparison to MobilenetV1.It uses 2 times fewer operations, has
higher accuracy, needs 30 percent fewer parameters and is about 30-40 percent
faster.
Also the non-linearities present in mobilenetv1 are removed.

Architecture of MobileNetV2 :

MobileNetV2 builds upon the ideas from MobileNetV1 , using depthwise

separable convolution as efficient building blocks. However, V2 introduces two new
features to the architecture:

1) linear bottlenecks between the layers.

2) shortcut connections between the bottlenecks.

The intuition is that the bottlenecks encode the model’s intermediate inputs and
outputs while the inner layer encapsulates the model’s ability to transform from
lower-level concepts such as pixels to higher level descriptors such as image
categories. Finally, as with traditional residual connections, shortcuts enable faster
training and better accuracy.

In MobileNetV2, there are two types of blocks. One is residual block with stride of
1 another one is block with stride of 2 for downsizing.
But their are three unique layers for both blocks
1. The first layer is 1×1 convolution with ReLU6.

2. The second layer is the depthwise convolution.

3. The third layer is another 1×1 convolution but without any non-linearity

Overall Architecture of Both MobileNetV1 and MobileNetV2 :

MobileNetV1
MobileNetV2