DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs
Liang-Chieh Chen, George Papandreou, Iasonas Kokkinos, Kevin Murphy, and Alan L. Yuille L.-C. Chen, G. Papandreou, and K. Murphy are with Google Inc. I. Kokkinos is with University College London. A. Yuille is with the Departments of Cognitive Science and Computer Science, Johns Hopkins University. The first two authors contributed equally to this work.
Abstract
In this work we address the task of semantic image segmentation with Deep Learning and make three main contributions that are experimentally shown to have substantial practical merit. First, we highlight convolution with upsampled filters, or ‘atrous convolution’, as a powerful tool in dense prediction tasks. Atrous convolution allows us to explicitly control the resolution at which feature responses are computed within Deep Convolutional Neural Networks. It also allows us to effectively enlarge the field of view of filters to incorporate larger context without increasing the number of parameters or the amount of computation. Second, we propose atrous spatial pyramid pooling (ASPP) to robustly segment objects at multiple scales. ASPP probes an incoming convolutional feature layer with filters at multiple sampling rates and effective fields-of-views, thus capturing objects as well as image context at multiple scales. Third, we improve the localization of object boundaries by combining methods from DCNNs and probabilistic graphical models. The commonly deployed combination of max-pooling and downsampling in DCNNs achieves invariance but has a toll on localization accuracy. We overcom
中文速览
图像语义分割（semantic image segmentation）长期面临三大难题：下采样导致特征图分辨率损失、物体尺度多变难以统一处理、以及卷积网络的平移不变性造成边界定位模糊。DeepLab针对这三个痛点分别提出了对应方案：用"空洞卷积"（atrous convolution，即在卷积核中插入空隙）在不增加参数量的前提下扩大感受野、保持高分辨率特征；用"空洞空间金字塔池化"（ASPP）在多个采样率上并行卷积，同时捕捉不同尺度的目标；最后在网络输出后接一个全连接条件随机场（CRF）对边界细节做精细修正。实验结果显示，DeepLab在PASCAL VOC 2012测试集上达到79.7% mIOU，并在另外三个数据集上也刷新了当时的最优成绩。这项工作的意义在于用简洁的两阶段级联（DCNN+CRF）就取得了极具竞争力的表现，所提出的空洞卷积和ASPP此后成为语义分割领域最广泛采用的基础组件。
原文 arXiv:1606.00915;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1606.00915v2