Should You Start with Machine Learning or Deep Learning for Computer Vision?

If you are interested in learning Computer Vision, it is generally advisable to begin with a course on Machine Learning (ML) before diving into Deep Learning (DL). This article will discuss why starting with ML is beneficial and the suggested learning path.

Foundational Knowledge

One of the primary reasons to start with Machine Learning is to gain foundational knowledge. A ML course introduces you to fundamental concepts such as algorithms, data preprocessing, feature extraction, and model evaluation. These basics are crucial for understanding the more complex architectures used in deep learning.

Traditional Algorithms

Many traditional ML algorithms like Support Vector Machines (SVMs), decision trees, and k-nearest neighbors are still relevant in Computer Vision tasks, especially for smaller datasets or simpler problems. Knowing how these algorithms work will give you a broader perspective and a solid base to build upon.

Deep Learning as a Subset

Deep Learning is a subset of Machine Learning that focuses on neural networks with many layers. Once you have a solid grasp of ML principles, transitioning to DL will be easier. You will understand how neural networks fit into the larger picture and be better prepared to learn about Convolutional Neural Networks (CNNs) and other advanced deep learning techniques.

Computer Vision Techniques

Many foundational techniques in Computer Vision can be understood through ML concepts. This will help you appreciate how deep learning models, such as CNNs, improve upon traditional methods. Understanding these concepts will make the learning process smoother and more effective.

Suggested Learning Path

To build a strong foundation for Computer Vision, follow this suggested learning path:

Start with Machine Learning

Look for a comprehensive ML course that covers both theory and practical applications. Some popular choices include the University of Washington’s Machine Learning course on Coursera, the University of Colorado Boulder’s Machine Learning course on edX, or the Georgia Institute of Technology’s Introduction to Machine Learning course.

Move to Deep Learning

Once you are comfortable with ML principles, take a course focused on Deep Learning, particularly one that emphasizes applications in Computer Vision. Some recommended courses include Andrew Ng’s NanoDegree in Deep Learning and the Deep Learning specialization offered by the University of London.

Explore Computer Vision

After grasping DL, dive into specific Computer Vision courses that apply what you have learned about deep learning to tasks such as image classification, object detection, and segmentation. Courses like the University of Washington’s Computer Vision specialization on Coursera can be excellent resources.

Conclusion

Following a structured learning path from ML to DL to Computer Vision will help you build a strong foundation and effectively learn these complex topics. While deep learning is the de facto method for solving Computer Vision and Natural Language Processing (NLP) problems, other ML techniques, such as decision trees, are still valuable in certain scenarios. Depending on your application and goals, using prebuilt services or learning from scratch can be beneficial.

Keywords

Machine Learning, Deep Learning, Computer Vision