基于多尺度卷积神经网络的场景标记

Scene Labeling Based on Multiscale Convolutional Network

场景标记是将图片中的像素按照其所属景物的种类来识别并进行标记。传统学习算法将训练集图片和某种学习机制相结合,利用后者的特点来提高训练正确率。提出一种基于多尺度卷积神经网络训练已知图像及其标记的方法,用测试集图片来验证其标记正确率。通过在Ubuntu系统上搭建快速机器学习环境Torch7来实现图片像素的场景标记。

Scene labeling is a method which we label each pixel in an image with the category of the object it belongs to. The traditional learning algorithms combine the family of images with some method which is used to improve accuracy of training. Presents a method that uses a multiscale convolution network trained from pixels with label known and gets verified by the test set of graph. The system is built on Ubuntu by Torch7 which is a kind of sharp environment for machine learning.