Scribd
Upload
14 views

Image Classification Using Cnn

Uploaded by

l.scorpiolistin20
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
14 views

Image Classification Using Cnn

Uploaded by

l.scorpiolistin20
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 15

IMAGE CLASSIFICATION

USING CNN

Student Name: k nethaji


Department: Information Technology
AGENDA
• DOMAIN INTRODUCTION

• EXISTING SYSTEM

• PROPOSED SYSTEM

• PROPOSED ARCHITECTURE

• SCREENSHOTS

• CONCLUSION

• FUTURE ENHANCEMENT

• OUTCOME OF THIS IDEA

• REFERENCES
DOMAIN INTRODUCTION

 Image classification is the process of sorting images into


categories, like recognizing animals or objects in a picture.
Convolutional Neural Networks (CNNs) are a type of deep
learning model that are very good at analyzing images. They
work by breaking images into small parts, finding patterns like
edges or shapes, and using these to classify the image.

 CNNs are widely used in tasks like detecting faces, recognizing


handwriting, or even diagnosing diseases from medical images.
They are powerful because they learn these patterns
automatically, making image classification more accurate and
efficient.
EXISTING SYSTEM

 Traditional systems used manual methods to find patterns in images, like


detecting edges or colors. These patterns were then used by basic machine
learning models to classify the images. However, these systems had
problems like low accuracy, difficulty handling complex images, and
requiring a lot of manual work. They also couldn’t adapt well to new types
of images. CNNs were introduced to solve these issues by automatically
learning patterns and improving accuracy
 Before CNNs, traditional image classification systems used algorithms
like:
 Support Vector Machines (SVM)
 K-Nearest Neighbors (KNN)
 Decision Trees
 Naive Bayes.
 These methods relied heavily on manually designed features, which made
them less accurate and less flexible compared to CNNs.
PROPOSED SYSTEM

 The proposed system uses CNNs to automatically learn


patterns from images, like edges and shapes. It provides:
 Better Accuracy for classifying images, even complex ones.
 Automatic Learning, removing the need for manual work.
 Faster Performance, handling large datasets easily.
 Flexibility to work on tasks like face recognition or object
detection.
 This makes it more powerful and efficient than older methods
PROPOSED ARCHITECTURE
SCREENSHOTS
Outcome of this Idea

In conclusion, we have discussed the challenges


faced by visually impaired individuals and how
Tensor Flow AI tool can help improve their quality
of life. With its ability to recognize objects and
read text, this tool has the potential to
revolutionize the way visually impaired
individuals navigate the world.
We have also discussed the benefits of using
Tensor Flow AI tool, including increased
independence and improved access to
information. While there may be challenges in
implementing this tool, it is important to
recognize its potential and work towards
overcoming these challenges.
As a society, it is our responsibility to ensure
that all individuals have equal access to
CONCLUSION

 In this project, we created a neural network using TensorFlow and Keras to


classify images from the Fashion MNIST dataset. The model consisted of an
input layer that flattened 28x28 pixel images, a hidden layer with 128 neurons
and ReLU activation to learn features, and an output layer with 10 neurons using
softmax to predict one of the 10 clothing categories.
 After training the model, we evaluated its performance on a test dataset. The
model achieved good accuracy in classifying the images and we visualized its
predictions to see how well it performed. The visualizations helped in
identifying correct and incorrect predictions.
 This simple neural network serves as a good foundation for image classification
tasks. Future improvements could involve using more advanced models, like
convolutional neural networks (CNNs), to achieve even better performance and
accuracy.
FUTURE ENHANCEMENT

 One way to improve the model is by using Convolutional Neural Networks


(CNNs). These are better at handling images because they can detect patterns in
images more effectively, which would make the model more accurate.
 Another improvement is data augmentation. This involves changing the
images in small ways, like rotating or flipping them, to make the model learn
from more variety. This can help prevent the model from becoming too
specialized to the training data and improve its performance on new images.
 Finally, we could make the model work in real-time, where it can classify
images immediately as they are uploaded or taken. This would be useful for
applications like online stores or security systems where fast results are needed.
 These improvements would make the model more accurate, faster, and ready
for use in real-world situations.
REFERENCES

 "Hands-On Machine Learning with Scikit-Learn,


Keras, and TensorFlow" by Aurélien Géron
 "Pattern Recognition and Machine Learning" by
Christopher M. Bishop.
 "Deep Learning" by Ian Goodfellow, Yoshua
Bengio, and Aaron Courville
 "Neural Networks and Deep Learning" by Michael
Nielsen
THANK
YOU

You might also like