Author: Zen and the Art of Computer Programming
1 Introduction
What is computer vision?
Computer Vision (CV) refers to the study of how computers can capture, analyze, and process information from various inputs (such as images, videos), and display it in a human-understandable form. It includes multiple subfields such as target detection, image segmentation, image tracking, image stylization, and face recognition. Its development began in the 1960s, has experienced many revolutions, and has become a popular direction in today's computer technology applications.
Why study computer vision?
As a programmer, you should have a certain understanding of computer vision technology and its related algorithms. If you want to master image processing, analysis, statistics, machine learning and other skills, then computer vision is an essential part. In addition, if your work involves knowledge of image processing and computer vision, you will benefit a lot.
What is Pytorch?
PyTorch is an open source Python library that provides a unified interface for building and training neural networks. It can be used for data loading, model definition, optimizer configuration, loss function definition, etc. Pytorch supports GPU acceleration, which allows us to use GPU computing to increase speed and achieve faster model training. Therefore, this series of tutorials on image classification based on Pytorch will focus on the powerful framework of Pytorch.
2. Explanation of basic concepts and terms
1. Image classification task
Given an image to be classified, our goal is to distinguish which category the image belongs to. There are generally two strategies for image classification in the field of computer vision:
- Classification based on manual features: This method is often called "template matching" or "pattern classification". It requires designing some rules or templates, and then finding the positions of these templates in the entire image.