课程: Self-Supervised Machine Learning

Self-supervised learning in computer vision

课程: Self-Supervised Machine Learning

开始一个月试用 购买团队方案

Self-supervised learning in computer vision

“

- [Instructor] Self-supervised learning techniques can be applied on any kind of data, but recently there has been a lot of research into self-supervised learning in the field of computer vision and that's what we'll focus on for the rest of this course. Let's see how self-supervised tasks can be helpful in this field. A common downstream task in the field of computer vision can be image classification, object detection, boundary mapping, et cetera. The self-supervised pretext task needs to learn visual representations of data, where the data is image, or maybe videos. Pretext tasks in computer vision are set up in such a way that they use observed features to try and predict hidden data or some property of the hidden data. Now, what are some pretext tasks in computer vision? Let's break this down. They can be pretext tasks for images, pretext tasks to use with videos and pretext tasks that can be used with video as well as sound. Let's start with images. There are a whole variety of pretext tasks that you can set up which can learn using self-supervised learning techniques. For example, you can predict image rotation. You can predict image colorization. You can predict inpainting of images. You can predict the relative position of objects or you can have the jigsaw task. Never mind if all of these tasks are not completely obvious when you first hear what they do. We'll be discussing them in detail over the next few videos.

内容