Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
image_classification.pptx
1. Image classification using CNN
Compiler: Parham Abolghasemi
Professor: Hamirdeza Bolhasani
2022, December 22
2. What Is Image Classification?
• It’s the task of extracting and categorizing useful information from a multi-band
raster image.
• The image is classified according to its visual content.
• The classification aim is to categorize all pixels of an image into one class.
3. Convolutional Neural Networks (CNNs)
• A branch of deep learning
• CNNs take a biological inspiration from the visual cortex which has a has small regions of cells that
are sensitive to specific regions of the visual field.
5. What we see Vs. What Computer Sees
For humans, this task is one of the first skills we learn and
it comes naturally and effortlessly as adults.
Being able to quickly recognize patterns, generalize from
prior knowledge, and adapt to different image
environments are difficult tasks for machines.
7. Closer Look 👀
• Scan the image
• Determine 0 & 1 pixels
• Image converts to 1D array
• Applying math operations
• ApplyingActivation function
• Classify the image
8. RELU Activation Function
• An operation called Rectified Linear Unit (RELU) has been used after every Convolution operation.
• It’s an element wise operation (applied per pixel) and replaces all negative pixel values in the feature map by zero.
• The purpose is to introduce non-linearity to the network.
9. CIFAR Dataset
• The CIFAR-10 dataset is a collection of images that are commonly used to train machine learning and computer vision algorithms. It
is one of the most widely used datasets for machine learning research.
• The CIFAR-10 dataset contains 60,000 color images in 10 different classes. (There are 6,000 images of each class)
A. Airplanes
B. Cars
C. Birds
D. Cats
E. Deer
F. Dogs
G. Frogs
H. Horses
I. Ships
J. Trucks
cifar-dataset
14. Normalizing Image
• the process that changes the range of pixel values.
• The purpose of Normalization is to bring image to range that is normal to sense.
• Channels are RGB
• Each pixel exists between 0 – 255
• Dividing return values to 255