文章大纲
数据增强Data Augmentation
在本文中,我们用规则数据和增强数据训练卷积神经网络。本文的目的是证明增强数据提高了泛化性能。
we will train a Convolutional Neural Network with Regular data and Augmented data. The purpose of this article is to show that the Augmented data improves generalization performance.
想象一个场景,无人机必须拍摄物体的照片。无人机在移动,物体也可能在移动。当拍摄图像时,我们并不总是会得到完美的图像。对象可能没有在图像中完全居中,或者对象可能在图像中旋转。在这种情况下,在完全居中或旋转的图像上训练的模型表现不佳。这就是为什么我们在旋转数据上训练一个模型,这样它可以在不完美的图像上表现良好。
Think of a scenario where a drone has to take a picture of an object. The drone is moving and the object can also possibly be moving. When an image is taken we arent always going to get perfect images. The subject may not be perfectly centered in the image or the subject may be r