Most of the proposed person re-identification algorithms conduct supervised training and testing on single labeled datasets with small size, so directly deploying these trained models to a large-scale real-world camera network may lead to poor performance due to underfitting. It is challenging to incrementally optimize the models by using the abundant unlabeled data collected from the target domain. To address this challenge, we propose an unsupervised incremental learning algorithm, TFusion, which is aided by the transfer learning of the pedestrians' spatio-temporal patterns in the target domain. Specifically, the algorithm firstly transfers the visual classifier trained from small labeled source dataset to the unlabeled target dataset so as to learn the pedestrians' spatial-temporal patterns. Secondly, a Bayesian fusion model is proposed to combine the learned spatio-temporal patterns with visual features to achieve a significantly improved classifier. Finally, we propose a learning-to-rank based mutual promotion procedure to incrementally optimize the classifiers based on the unlabeled data in the target domain. Comprehensive experiments based on multiple real surveillance datasets are conducted, and the results show that our algorithm gains significant improvement compared with the state-of-art cross-dataset unsupervised person reidentification algorithms.
IP-based TV systems are widely used to stream video contents on the Internet. Compared with the traditional broadcast TV systems, IP-based TV systems suffer from much longer channel switching delays. In this paper, we propose a new IP-based streaming framework, called FIPTV (Fast IP-based TV), to achieve close-to-zero channel switching delay at the price of extra download bandwidth and increased playback lags. In FIPTV, other than the channel being watched, a client also downloads an extra combination virtual channel, called BUS (Backing United Stream), which consists of video segments sequentially sampled from a set of target channels that a client might switch to in the near future.Video segments downloaded from the combination channel will be cached in a local buffer. When the client issues a channel switch request to a target channel, the client will immediately playback the most recently downloaded video of the target channel, leading to close-to-zero channel switching time, but a positive playback lag. Through analysis and simulations, we show that short average playback lag can be achieved cross all channels through carefully designed channel scheduling algorithms on the BUS channel by considering channel popularity. We implement the proposed streaming framework in real systems. Through experiments on the Internet, we show that the actual channel switching delay can be reduced to less than 0.25 seconds, which is much shorter than that of the popular Internet video streaming services.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.