Paper reading notes: Fully convolutional Networks for Semantic segmentation

This is CVPR 2015 to get the best paper candidate paper.

Paper: Fully convolutional Networks for Semantic segmentation

Respect the original, reproduced please specify: http://blog.csdn.net/tangwei2014

1. Overview & Key contributions

This paper presents a end-to-end method of semantic segmentation, referred to as FCN.

As shown, directly take segmentation's ground truth as the supervisory information, train an end-to-end network, let the network do pixelwise prediction, directly predict the label map.

2. Question & Workaround 1) How to do pixelwise prediction?

The traditional network is subsampling, the corresponding output size will be reduced, to do pixelwise prediction, the output size must be guaranteed.

Workaround:

(1) The final fully connected layer of the traditional network, such as Alexnet,vgg, becomes the convolution layer.

For example, the first fully connected layer in VGG16 is 25088x4096, which is interpreted as a convolution core of 512x7x7x4096, then if the convolution operation is performed on a larger input image (the lower half), the original output 4096-D feature node (the upper part), It will output a coarse feature map.

The advantage of this is that you can take good advantage of the trained supervised pre-training network, do not like the existing methods, from beginning to end training, only need to fine-tuning, training efficient.

(2) Add In-network upsampling layer.

The feature map obtained in the middle is sampled on bilinear, which is the deconvolution layer. The implementation of the conv of the forward and reverse transfer process can be reversed.

2) How to refine and get better results?

Upsampling the step is 32, the input is 3x500x500, the output is 544x544, the edge is not good.

Workaround:

By using the Skip layer method, the step of upsampling is reduced at the shallow layer, multiple label map prediction are obtained, and then fusion is done.

3. Results

Of course it's state-of-the-art. Feel it:

Copyright NOTICE: This article for Bo Master original article, without Bo Master permission not reproduced.

Paper reading notes: Fully convolutional Networks for Semantic segmentation