On deeply learning features for automatic person re-identification

The automatic person re-identification problem resides in matching an unknown person image to a database of previously labeled images of people. Comparison among two image features is commonly accomplished by distance metrics. Although features and distance metrics can be handcrafted or trainable the latter type has demonstrated more potential to breakthroughs in achieving state-of-the-art performance over public data sets. A recent paradigm that allows to work with trainable features is deep learning. In this book we present a novel deep learning strategy so called coarse-to-fine learning (CFL) as well as a novel type of feature - the convolutional covariance features (CCF) for person re-identification. CFL is based on the human learning process. After extracting the convolutional features via CFL those ones are then wrapped in covariance matrices composing the CCF. The performance of the proposed framework was assessed comparatively against 18 state-of-the-art methods by using public data sets (VIPeR i-LIDS CUHK01 and CUHK03) achieving superior performance.