一、回顾《Fully Convolutional Networks for Semantic Segmentation》
二、看了一篇新的论文
《An Improved Deep Learning Architecture for Person Re-identification》
1、输入一对图像,网络输出这对图像的相似度值,
新引入的网络层包括跨输入邻域差值层,根据图像对的卷积特征图计算局部关联,之后使用加和特征对输出特征图的邻域进行加和,最后计算远距离像素点的关联性
2、网络结构
two layers of tied convolution with max pooling
cross-input neighborhood differences,
patch summary features
across-patch features
higher-order relationships
softmax层估计输入的图像是否是同一个人
三、跑代码
1)跑《An Improved Deep Learning Architecture for Person Re-identification》
2)跑DCGAN的代码,为下一篇论文《Unlabeled Samples Generated by GAN Improve the Person Re-identification Baseline in vitro》做准备。
四、看论文《An Improved Deep Learning Architecture for Person Re-identification》源码
本文回顾了《FullyConvolutionalNetworksforSemanticSegmentation》,并深入解析了《AnImprovedDeepLearningArchitectureforPersonRe-identification》论文。该文介绍了用于行人重识别的改进深度学习架构,包括一对图像输入计算相似度值,网络层如跨输入邻域差值层、patchsummaryfeatures及across-patchfeatures等。此外,还提到了softmax层估计输入图像是否属于同一人,并分享了跑代码的经验。
2260





