Convolutional Feature Masking for Joint Object and Stuff Segmentation
Jifeng Dai Kaiming He Jian Sun Microsoft Research
Abstract
The topic of semantic segmentation has witnessed considerable progress due to the powerful features learned by convolutional neural networks (CNNs) [13]. The current leading approaches for semantic segmentation exploit shape information by extracting CNN features from masked image regions. This strategy introduces artificial boundaries on the images and may impact the quality of the extracted features. Besides, the operations on the raw image domain require to compute thousands of networks on a single image, which is time-consuming.
中文速览
图像语义分割（semantic segmentation）长期面临两大瓶颈：直接在原始图像上做区域掩码会引入训练时从未见过的人工边界，破坏特征质量；同时对成千上万个候选区域逐一跑网络，速度极慢。这篇论文提出了一种叫做"卷积特征掩码"（Convolutional Feature Masking, CFM）的方法：只对完整图像提取一次卷积特征图，然后把分割候选区域（超像素）投影到特征图上直接做掩码，从而绕开了在原始像素上裁图的做法，既保留了特征完整性，又大幅节省计算量。实验表明，该方法在 PASCAL VOC 2012 目标分割基准上达到了当时最优水平，使用 VGG-16 主干网络时分割精度达到 61.8%，比此前最强方法 SDS 快约 150 倍，并且是首个将深度学习应用于 PASCAL-CONTEXT 物体与"背景类别"联合分割任务的方法。这项工作的意义在于，它打通了高效特征复用（来自 SPP-Net）与精细形状感知之间的通道，为后续实例分割和全景分割研究提供了重要的技术基础。
原文 arXiv:1412.1283;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1412.1283v4