跳到主要内容
Learn Motion
探索课程产品能力价格

下载 Learn Motion

用手机扫描二维码,下载 App。

App Store

扫码从 App Store 下载 Learn Motion打开商店页面

Google Play

扫码从 Google Play 下载 Learn Motion打开商店页面
登录
Learn Motion

How AI Sees: A Conceptual Guide to Image and Video Understanding

1What a Machine Actually Sees2From Pixels to Patterns: Features3How Deep Networks Learn to See4Beyond Labels: Locating and Describing What Is Seen5Adding Time: Understanding Video6How These Systems Learn and How We Judge Them
Adding Time: Understanding Video

Optical flow: turning movement into a field of arrows

2 / 5
Watch the two frames slide past each other. For each patch, the animation shows where that patch went, and draws an arrow for the displacement. Collect every arrow and you have the flow field. Notice the arrows are a measurement of the picture changing, not of the objects moving. If the camera pans across a still wall, the arrows are large even though nothing in the scene moved. That is why we call it apparent motion — it tracks what the image is doing, which usually, but not always, matches what the world is doing.
0:00 / 0:00

Optical flow is the apparent motion of brightness between two consecutive frames. For each small patch of the image, the method asks: where did this patch move to in the next frame? The answer is a displacement — a horizontal and a vertical shift — and collecting that answer for every patch produces a dense field of arrows, one per location.

The arrows are not the objects themselves. They are a separate measurement of how the picture is changing. A camera panning across a still wall produces large flow arrows even though nothing in the scene moved; a person walking toward a stationary camera produces arrows radiating outward from the center. This is why flow is described as apparent motion rather than true motion: it captures what the image is doing, which usually but not always matches what the world is doing.

Flow is useful because it strips away appearance and keeps only change. Two very different-looking actions — a bird flapping and a flag waving — can produce similar flow patterns, and two similar-looking scenes can produce very different flow. That complementarity is what makes flow a useful second source of evidence alongside the raw frames.

上一页2 / 5下一页

Learn Motion

生成课程,认真学会。

运营方:Wuhan Daoyin Technology Co., Ltd.

联系我们: [email protected]
隐私政策用户服务协议

© 2026 Learn Motion