2. WHAT IS RESNET ? ● ResNet, short for Residual Networks, is a classic
neural network used as a backbone for many
computer vision applications.
● ResNet makes it possible to train up to hundreds or even
thousands of layers and still achieves compelling
performance.
Why Use
ResNet?
● Humans can gather a wide variety of information
from an image. our goal in computer vision is to
make the same possible with computers or
machines.
● A normal artificial neural network was found insufficient for
such processing when there is a need to gather complex
information
3. A picture can be considered as a matrix of numbers with each element
representing a pixel value ( values from 1-225 ), which depicts the
amount of a particular color.
Why Use
ResNet?
A colored image is a stack of such
red, green, and blue layers.
The values of these pixels are fed into
the neural network as inputs
But when the quality of the
image increases
number of pixels increases
which in turn increases the
inputs nodes and this
leads to overfitting.
Adding the pixel values directly into the neural
network can also cause loss of spatial data,
4. Convolutional layers are used to
convolve the image into a smaller
size using filters or kernels
So we use convolutional neural networks
To train a NN to identify fine details in an image we need to have more
nodes or neurons but when increasing the number of nodes or neurons
in a single layer it can cause overfitting. So we instead increase the
number of layers with these nodes.
5. But very deep neural networks are difficult to train because of vanishing
and exploding gradients Problems
Resnet uses skip connections which
allows take the activation from one
layer and feed it to another layer
much deeper in the neural network
using this deeper neural network
with even over 100 layers can be
trained
Ordinary
Y=F(X)
Y=f(x)+x