BoxSup: Exploiting Bounding Boxes to Supervise Convolutional Networks for Semantic Segmentation
Jifeng Dai Kaiming He Jian Sun Microsoft Research
Abstract
Recent leading approaches to semantic segmentation rely on deep convolutional networks trained with human-annotated, pixel-level segmentation masks. Such pixel-accurate supervision demands expensive labeling effort and limits the performance of deep networks that usually benefit from more training data. In this paper, we propose a method that achieves competitive accuracy but only requires easily obtained bounding box annotations. The basic idea is to iterate between automatically generating region proposals and training convolutional networks. These two steps gradually recover segmentation masks for improving the networks, and vise versa. Our method, called “BoxSup”, produces competitive results (e.g., 62.0% mAP for validation) supervised by boxes only, on par with strong baselines (e.g., 63.8% mAP) fully supervised by masks under the same setting. By leveraging a large amount of bounding boxes, BoxSup further unleashes the power of deep convolutional networks and yields state-of-the-art results on PASCAL VOC 2012 and PASCAL-CONTEXT [24].
中文速览
像素级标注太贵、太费力，训练语义分割（semantic segmentation）模型时能不能改用更容易获取的边界框（bounding box）标注？BoxSup 方法给出了肯定的答案：它交替执行两步——先用无监督区域候选算法（如 MCG）从边界框中自动生成近似分割掩膜，再用这些掩膜监督卷积网络训练，更新后的网络反过来帮助挑选更准确的候选区域，如此迭代循环，让粗糙的初始掩膜越来越精准。实验表明，仅用边界框监督可在 PASCAL VOC 2012 上达到 62.0% mIoU，与使用完整像素掩膜的强基线（63.8%）相差无几；进一步引入大规模现有边界框数据后，BoxSup 甚至刷新了该基准及 PASCAL-CONTEXT 上的最优结果。这项工作意义在于大幅降低了语义分割的数据标注门槛，同时证明了廉价的框级标注同样能持续驱动深度网络性能提升。
原文 arXiv:1503.01640;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1503.01640v2