A Brief Brush-up On Deep Learning Concepts- Introduction

So…

I was preparing for the AI engineer role as a fresher and I thought, I would document it so that it might help others also who want last minute revision on deep learning concepts.

Deep learning is one of the technologies under Artificial Intelligence which is being used by so many industries and has been growing at a high speed.It is such a trendy word and as the name says the depth of the concept is infinite.I am starting with the basic stuff in deep learning.Image processing is one of the applications of deep learning.Let’s go over some important topics in Image processing.

Photo by Hitesh Choudhary on Unsplash

In this article you will revise 3 topics in Image processing applications.They are: 1)Image classification 2)Object detection 3)Image segmentation

Image classification

As a beginner to deep learning and Image learning one has to first learn what CNN is.The convolutional neural network (CNN) is a class of deep learning neural networks. CNNs represent a huge breakthrough in image recognition since they are fast and accurate.The image classification is used everywhere from facebook to high-tech robots.There can be 2 or more different class labels or categories.There are 3 types of classification: binary,multi-class and multi-label.

How are they working?

A CNN has

• Convolutional layers
• ReLU layers
• Pooling layers
• A Fully connected layer

An usual CNN architecture would look something like this:

Input ->Convolution-->ReLU ->Convolution ->ReLU ->Pooling ->
ReLU ->Convolution ->ReLU ->Pooling ->Fully Connected

A detailed procedure of how cnn works will be explained in the next article of this series.

Object detection

Locate the presence of objects with a bounding box and types or classes of the located objects in an image.

• Input: An image with one or more objects, such as a photograph.
• Output: One or more bounding boxes and a class label for each bounding box.

There can be single object detection or multi-object detection.A computer vision technique is used to propose candidate regions or bounding boxes of potential objects in the image called “selective search,” although the flexibility of the design allows other region proposal algorithms to be used.We use usually YOLO or RCNN architectures for this procedure.

Image segmentation

image from aws website

How does image segmentation work?

We divide or partition the image into various parts called segments. It’s not a great idea to process the entire image at the same time as there will be regions in the image which do not contain any information. By dividing the image into segments, we can make use of the important segments for processing the image. That, in a nutshell, is how image segmentation works.

An image is a collection or set of different pixels. We group together the pixels that have similar attributes using image segmentation. Take a moment to go through the below visual (it’ll give you a practical idea of image segmentation).

There are 2 types of segmentation:

1. Semantic segmentation:semantic segmentation will focus on classifying all objects which belong to same label as a single instance.
2. Instance segmentation:Instance segmentation, on the other hand. will identify each of these objects individually.

Region-based segmentaion:

1. If we want to divide the image into two regions (object and background), we define a single threshold value. This is known as the global threshold.
2. 2.If we have multiple objects along with the background, we must define multiple thresholds. These thresholds are collectively known as the local threshold.

This article is just the beginning of our journey to learn all about image processing in deep learning. In the coming articles, we will deep dive into the complete procedure of each of these topics. So stay tuned!