- low-level feature:通常是指圖像中的一些小的細節信息，例如邊緣(edge)，角(corner), 顏色(color),像素(pixels),梯度(gradients)等，這些信息可以通過濾波器、SIFT或HOG獲取；
- hight-level feature:是建立在low level feature之上的，可以用於圖像中目標或物體形狀的識別和檢測，具有更豐富的語義信息。
通常卷積神經網絡中都會使用這兩種類型的features:卷積神經網絡的前幾層學習Low level feature, 后幾層學習的是high level feature.
Quora上面也有這么一段解釋：
Low-level features are minor details of the image, like lines or dots, that can be pickup by , say, a convolutional filter (for reaaly low-level things) or SIFT or HOG (for more abstract things like edges).
High levle features are built on top of low-level features to detect objects and shapes in the image.