The document is a technical presentation on Convolutional Neural Networks (CNNs), detailing their background, operation, and features. It discusses the advantages and drawbacks of CNNs, such as their effectiveness in pattern recognition and the high resource requirements for training. Additionally, it covers various applications of CNNs in fields like healthcare, automotive, and social media.

1. 1
Sharad Institute Of Technology
College Of Engineering Yadrav
Seminar
On
Introcution To CNN
Submitted To: Submitted By:
Miss. P. B. Shinde Vidya Vijay Mali

2. A Technical Presentation on
Presented By:
Vidya Vijay Mali
2
Introductionto
CNN

3. Contents
3
Background of CNNs
What Is a CNN?
How does it work?
What Is a Pooling Layer?
CNNs vs. neural networks
Benefits of using CNNs
Drawbacks of CNN
Application

4. Background of CNNs
4 8 March 2015
 first developed and used around the 1980s
 It was mostly used in the postal sectors to read zip
codes, pin codes, digit recognition.
 CNNs were only limited to the postal sectors and it
failed to enter the world of machine learning.
 Because it requires a large amount of data to train
and also requires a lot of computing resources.

5. What is CNN?
5

6.  CNN/ConvNet
 class of deep neural networks, most commonly applied to
analyze visual imagery
 It uses a special technique called Convolution.
 in mathematics convolution is a mathematical operation on two
functions that produces a third function that expresses how the
shape of one is modified by the other.
6

7. 9
How does it work?
Fig. RGB Image matrix

9. Fig. Gray colo rimage convolution

11. In the case of RGB color, channel take a look at this animation to understand its working

12. 12
CNN work flow

13. What Is a Pooling Layer?
 responsible for reducing the spatial size of the
Convolved Feature.
 to decrease the computational power required to
process the data by reducing the dimensions.
 two types of pooling
Average pooling
Max pooling

14. Max poolig Average poolig

15. Advantages
Excellent in spotting patterns and characteristics in
signals like as photos, movies, and sounds.
Resistant to scaling, rotation, and translation invariance.
There's no need for manually extracting features with
end-to-end training.
Can attain excellent accuracy while handling massive
amounts of data.

16. Drawbacks
Expensive to train computationally and memory-
intensive.
Insufficient data or improper regularisation might lead to
overfitting.
Need a lot of data that has been tagged.
Limited interpretability makes it challenging to
comprehend what the network has discovered

17. Applications
Healthcare
Automotive
Social media
Facial recognition
Audio processing for virtual assistants

18. 18

19. 19