Object Contour Detection with a Fully Convolutional Encoder-Decoder Network

CVPR 2016 使用 CNN 网络进行 物体轮廓提取

Object Contour Detection with a Fully Convolutional Encoder-Decoder Network

我们将轮廓检测问题看做一个图像二值标记问题。
We formulate contour detection as a binary image labeling problem where “1” and “0” indicates “contour” and“non-contour”, respectively

Object Contour Detection with a Fully Convolutional Encoder-Decoder Network_第1张图片

CNN网络模型如下:

Object Contour Detection with a Fully Convolutional Encoder-Decoder Network_第2张图片

网络的 encoder 部分直接使用 VGG模型对应的部分,然后使用 decoder还原尺寸大小。
Object Contour Detection with a Fully Convolutional Encoder-Decoder Network_第3张图片

这里我们使用了文献【28】对轮廓真值数据进行微调整
Object Contour Detection with a Fully Convolutional Encoder-Decoder Network_第4张图片

结果很好:

Object Contour Detection with a Fully Convolutional Encoder-Decoder Network_第5张图片

Object Contour Detection with a Fully Convolutional Encoder-Decoder Network_第6张图片

你可能感兴趣的:(ZJ,CVPR,2016)