Recurrent Neural Networks (RNNs) are powerful sequence modeling tools. However, when dealing with high dimensional inputs, the training of RNNs becomes computational expensive due to the large number of model parameters. This hinders RNNs from solving many important computer vision tasks, such as Action Recognition in Videos and Image Captioning. To overcome this problem, we propose a compact and flexible structure, namely Block-Term tensor decomposition, which greatly reduces the parameters of RNNs and improves their training efficiency. Compared with alternative low-rank approximations, such as tensortrain RNN (TT-RNN), our method, Block-Term RNN (BT-RNN), is not only more concise (when using the same rank), but also able to attain a better approximation to the original RNNs with much fewer parameters. On three challenging tasks, including Action Recognition in Videos, Image Captioning and Image Generation, BT-RNN outperforms TT-RNN and the standard RNN in terms of both prediction accuracy and convergence rate. Specifically, BT-LSTM utilizes 17,388 times fewer parameters than the standard LSTM to achieve an accuracy improvement over 15.6% in the Action Recognition task on the UCF11 dataset.
A video recording of an examination by Wireless Capsule Endoscopy (WCE) may typically contain more than 55,000 video frames, which makes the manual visual screening by an experienced gastroenterologist a highly time-consuming task. In this paper, we propose a novel method of epitomized summarization of WCE videos for efficient visualization to a gastroenterologist. For each short sequence of a WCE video, an epitomized frame is generated. New constraints are introduced into the epitome formulation to achieve the necessary visual quality for manual examination, and an EM algorithm for learning the epitome is derived. First, the local context weights are introduced to generate the epitomized frame. The epitomized frame preserves the appearance of all the input patches from the frames of the short sequence. Furthermore, by introducing spatial distributions for semantic interpretation of image patches in our epitome formulation, we show that it also provides a framework to facilitate the semantic description of visual features to generate organized visual summarization of WCE video, where the patches in different positions correspond to different semantic information. Our experiments on real WCE videos show that, using epitomized summarization, the number of frames have to be examined by the gastroenterologist can be reduced to less than one-tenth of the original frames in the video.
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.