Convolutional Neural Networks (CNN) in Deep Learning

8 min read

Since the 1950s, AI researchers have worked on systems to understand visual data, leading to Computer Vision. In 2012, a breakthrough came with AlexNet, developed by Alex Krizhevsky at the University of Toronto. It achieved 85% accuracy in the ImageNet contest, far surpassing previous models. This success was driven by Convolutional Neural Networks (CNNs), which mimic human vision. CNNs are now vital for tasks like image classification, object detection, and segmentation. Modern CNNs use Python and advanced techniques to learn image features, relying on hyperparameters and optimization for training. Advances like VGG, ResNet, and EfficientNet have expanded CNN applications in fields like autonomous driving and medical imaging. In this article, you will get to know all about CNN, its advantages and disadvantages, and how CNN works.

What is a Convolutional Neural Network (CNN)?

In deep learning, a convolutional neural network (CNN/ConvNet) is a class of deep neural networks, most commonly applied to analyze visual imagery. The CNN architecture uses a special technique called Convolution instead of relying solely on matrix multiplications like traditional neural networks. Convolutional networks use a process called convolution, which combines two functions to show how one changes the shape of the other.

But we don’t need to go behind the mathematics part to understand what a CNN is or how it works. The bottom line is that the role of the convolutional networks is to reduce the images into a form that is easier to process, without losing features that are critical for getting a good prediction.

Read about the Guide to Deep Learning and Neural Networks

How Does CNN Work?

Before we go to the working of Convolutional neural networks (CNN), let’s cover the basics, such as what an image is and how it is represented. An RGB image is nothing but a matrix of pixel values having three planes, whereas a grayscale image is the same, but it has a single plane. Take a look at this image to understand more.

For simplicity, let’s stick with grayscale images as we try to understand how CNNs work.

The above image shows what a convolution is. We take a filter/kernel(3×3 matrix) and apply it to the input image to get the convolved feature. This convolved feature is passed on to the next layer.

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

The number of parameters in a CNN layer depends on the size of the receptive fields (filter kernels) and the number of filters. Each neuron in a CNN layer receives inputs from a local region of the previous layer, known as its receptive field. The receptive fields move over the input, calculating dot products and creating a convolved feature map as the output. Usually, this map then goes through a rectified linear unit (ReLU) activation function. Classic CNN architectures like LeNet and more modern ones like ResNet employ this fundamental principle.

Convolutional neural networks are composed of multiple layers of artificial neurons.

Artificial Neurons’ Role in CNN

Artificial neurons, a rough imitation of their biological counterparts, are mathematical functions that calculate the weighted sum of multiple inputs and output an activation value. When you input an image into a ConvNet, each layer generates several activation functions that are passed on to the next layer for feature extraction.

Check out this article about the Machine Learning Algorithms

Feature Extraction in CNN

The first layer usually extracts basic features such as horizontal or diagonal edges. This output is passed on to the next layer, which detects more complex features such as corners or combinational edges. As we move deeper into the network, it can identify even more complex features such as objects, faces, etc. Unlike recurrent neural networks, ConvNets are feed-forward networks that process the input data in a single pass.

Based on the activation map of the final convolution layer, the classification layer outputs a set of confidence scores (values between 0 and 1) that specify how likely the image is to belong to a “class.” For instance, if you have a ConvNet that detects cats, dogs, and horses, the output of the final layer is the possibility that the input data contains any of those animals. Gradient descent is commonly used as the optimization algorithm during training to adjust the weights of the input layer and subsequent layers.

Convolutional neural networks (CNNs) in Deep Learning

CNNs were first developed and used around the 1980s. The most that a Convolutional Neural Network (CNN) could do at that time was recognize handwritten digits. It was mostly used in the postal sector to read zip codes, PIN codes, etc. The important thing to remember about any deep learning model is that it requires a large amount of data to train and also requires a lot of computing resources. This was a major drawback for CNNs at that period, and hence, CNNs were only limited to the postal sectors and it failed to enter the world of machine learning. Backpropagation, the algorithm used to train neural networks, was also computationally expensive at the time.

• In 2012, Alex Krizhevsky recognized the potential to revive deep learning using multi-layered neural networks.
• This revival was made possible due to three key factors:
  1. The availability of large datasets.
  2. The creation of specialized datasets like ImageNet, which contained millions of labeled images.
  3. The increased availability of computing resources.
• These advancements allowed researchers to bring back and improve upon earlier approaches to deep learning.

Read More about deep learning in this article!

What is a Pooling Layer?

Similar to the Convolutional Layer, the Pooling layer is responsible for reducing the spatial size of the Convolved Feature. This is to decrease the computational power required to process the data by reducing the dimensions. There are two types of pooling: average pooling and max pooling. I’ve only had experience with Max Pooling so far I haven’t faced any difficulties.

So what we do in Max Pooling is we find the maximum value of a pixel from a portion of the image covered by the kernel. Max Pooling also performs as a Noise suppressor. It discards the noisy activations altogether and also performs denoising along with dimensionality reduction.

On the other hand, Average Pooling returns the average of all the values from the portion of the image covered by the Kernel. Average Pooling simply performs dimensionality reduction as a noise-suppressing mechanism. Hence, we can say that Max Pooling performs a lot better than Average Pooling.

Limitations of Convolutional Neural Networks (CNNs)

Despite the power and resource complexity of CNNs, they provide in-depth results. At the root of it all, it is just recognizing patterns and details that are so minute and inconspicuous that they go unnoticed to the human eye. But when it comes to understanding the contents of an image, it fails.

Let’s take a look at this example. When we pass the following image to a CNN, it detects a person in their mid-30s and a child, probably around 10 years old. But when we look at the same image, we start thinking of multiple different scenarios. Maybe it’s a father and son day out, a picnic, or maybe they are camping. Maybe it is a school ground, and the child scored a goal, and his dad is happy, so he lifts him.

These limitations are more than evident when it comes to practical applications. For example, CNNs were widely used to moderate content on social media. But despite the vast resources of images and videos that they were trained on, they still weren’t able to completely block and remove inappropriate content. As it turns out, it flagged a 30,000-year-old statue with nudity on Facebook.

• Studies indicate that CNNs (Convolutional Neural Networks) trained on datasets like ImageNet often struggle to detect objects when lighting conditions or viewing angles change.
• This raises questions about their effectiveness in varied real-world scenarios.
• Despite these limitations, CNNs have significantly advanced artificial intelligence.
• They are widely used in applications such as facial recognition, image search, photo editing, and augmented reality.
• Advances in CNNs have led to remarkable and practical achievements.
• However, these systems are still far from replicating the core aspects of human intelligence.

Advantages and Disadvantages of Convolutional Neural Networks (CNN)

Advantages

1. Automatically learn features without manual extraction.
2. Use shared weights, reducing parameters, and improving efficiency.
3. Recognize patterns regardless of their position in the input.
4. Capture both low-level and high-level features effectively.
5. Versatile and applicable to images, audio, video, and text.

Disadvantages

1. Require high computational power and resources.
2. Need large amounts of labeled data for training.
3. Difficult to interpret how decisions are made.
4. Can overfit, especially with small datasets.
5. Require fixed input sizes, limiting flexibility.

Conclusion

In this article, we’ve explored Convolutional Neural Networks (CNNs), delving into their functionality, background, and the role of pooling layers. Despite their effectiveness in image recognition, CNNs also come with limitations, including susceptibility to adversarial attacks and high computational requirements. CNNs are trained using a loss function that measures the difference between the predicted output and the ground truth. Fine-tuning pre-trained models on specific image data is a common practice to achieve better performance.

Additionally, CNNs can be used for segmentation tasks, which involve labeling each pixel in an image. Unlike traditional multilayer perceptrons, the network architecture of CNNs is designed to take advantage of the spatial and temporal dependencies in image data. Overall, CNNs have revolutionized the field of computer vision and continue to be an active area of research. Despite their effectiveness in image recognition, CNNs also come with limitations, including susceptibility to adversarial attacks and high computational requirements.

This article was published as part of the Data Science Blogathon.

Frequently Asked Questions

Login to continue reading and enjoy expert-curated content.

Free Courses

Become an Author

Share insights, grow your voice, and inspire the data community.

• Reach a Global Audience
• Share Your Expertise with the World
• Build Your Brand & Audience

• Join a Thriving AI Community
• Level Up Your AI Game
• Expand Your Influence in Genrative AI

👁 imag