DeepLab v2算法的简介(论文介绍)
DeepLabv2是DeepLabv1的改进版本,改进的不多,主要是用多尺度提取获得更好的分割效果。
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 overcome this by combining the responses at the final DCNN layer with a fully connected Conditional Random Field (CRF), which is shown both qualitatively and quantitatively to improve localization performance. Our proposed “DeepLab” system sets the new state-of-art at the PASCAL VOC-2012 semantic image segmentation task, reaching 79.7% mIOU in the test set, and advances the results on three other datasets: PASCAL-Context, PASCAL-Person-Part, and Cityscapes. All of our code is made publicly available online.
本文研究了基于深度学习的语义图像分割问题,并提出了三个具有实际应用价值的主要研究方向。首先,我们强调卷积与上采样滤波器,或“atrous卷积”,在密集预测任务中是一个强大的工具。Atrous卷积允许我们显式地控制在深度卷积神经网络中计算特征响应的分辨率。它还允许我们有效地扩大过滤器的视野,在不增加参数数量或计算量的情况下合并更大的上下文。其次,提出了一种基于空间金字塔池化 (ASPP)的多尺度鲁棒分割方法。ASPP使用多个采样速率的过滤器和有效的视图字段探测传入的卷积特征层,从而在多个尺度上捕获对象和图像上下文。第三,结合DCNNs方法和概率图形模型,改进了目标边界的定位。DCNNs中常用的最大池和下采样的组合实现了不变性,但对定位精度有一定的影响。我们通过将DCNN最后一层的响应与一个完全连接的条件随机场(CRF)相结合来克服这个问题,该条件随机场在定性和定量上都显示出来,以提高定位性能。我们提出的“DeepLab”系统在PASCAL VOC-2012语义图像分割任务中设置了新的技术状态,在测试集中达到了79.7%的mIOU,并在其他三个数据集:PASCAL-Context, PASCAL-Person-Part,和Cityscapes上提出了结果。我们所有的代码都在网上公开。
CONCLUSION
Our proposed “DeepLab” system re-purposes networks trained on image classification to the task of semantic segmentation by applying the ‘atrous convolution’ with upsampled filters for dense feature extraction. We further extend it to atrous spatial pyramid pooling, which encodes objects as well as image context at multiple scales. To produce semantically accurate predictions and detailed segmentation maps along object boundaries, we also combine ideas from deep convolutional neural networks and fully-connected conditional random fields. Our experimental results show that the proposed method significantly advances the state-ofart in several challenging datasets, including PASCAL VOC 2012 semantic image segmentation benchmark, PASCALContext, PASCAL-Person-Part, and Cityscapes datasets.
我们提出的“DeepLab”系统将训练有素的图像分类网络重新用于语义分割任务,利用带上采样滤波器的“atrous convolution”进行密集特征提取。我们进一步将其扩展到空间金字塔池,它在多个尺度上编码对象和图像上下文。为了产生精确的语义预测和沿着目标边界的详细分割地图,我们还结合了深度卷积神经网络和全连通条件随机域的思想。实验结果表明,该方法在PASCAL VOC 2012语义图像分割基准测试、PASCALContext, PASCAL-Person-Part和Cityscapes数据集等多个具有挑战性的数据集上都取得了显著的进步。
论文
Liang-ChiehChen, George Papandreou, IasonasKokkinos, Kevin Murphy, Alan L. Yuille.
DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, AtrousConvolution,
and Fully Connected CRFs.
IEEE Transactions on Pattern Analysis and Machine Intelligence ( Volume: 40 , Issue: 4 , April 1 2018 )应该是2017
https://arxiv.org/abs/1606.00915
0、实验结果
1、基于VGG-16的DeepLabmodel中,ASPP对PASCAL VOC 2012 valset性能(平均IOU)的影响。
Effect of ASPP on PASCAL VOC 2012 valset performance (mean IOU) for VGG-16 based DeepLabmodel.
LargeFOV: single branch, r = 12 .
ASPP-S: four branches, r= { 2, 4, 8, 12 } .
ASPP-L: four branches, r = { 6, 12, 18, 24 } .
多尺度+大感受野可显著提高语义分割效果
2、PASCAL VOC 2012 valresults输入图像和论文中的DeepLabresults之前/之后的CRF
PASCAL VOC 2012 valresults. Input image and our DeepLabresults before/after CRF
3、ASPP与基线LargeFOV模型进行定性分割
Qualitative segmentation results with ASPP compared to the baseline LargeFOV model.
采用多个大FOV的ASPP-L模型可以成功捕获多个尺度的目标和图像上下文。
4、PASCAL VOC 2012测试集性能
Performance on PASCAL VOC 2012 test set
在NVidia Titan X GPU 上运行速度达到了8FPS,全连接CRF 平均推断需要0.5s ,在耗时方面和DeepLab-v1无差异,但在PASCAL VOC-2012 达到79.7 mIOU。
1、DeepLab-v2 改进点
(1)、用多尺度特征提取获得更好的分割效果
目标存在多尺度的问题,DeepLabv1中是用多个MLP结合多尺度特征解决,虽然可以提升系统的性能,但是增加了特征计算量和存储空间。
受到SpatialPyramidPooling(SPP)的启发,提出了一个类似的结构,在给定的输入上以不同采样率的空洞卷积并行采样,相当于以多个尺度捕捉图像的上下文,称为ASPP(atrousspatialpyramidpooling)模块。
DeepLab v2算法的架构详解
更新……
DeepLab v2算法的案例应用
更新……