Generating temporal action proposals remains a very challenging problem, where the main issue lies in predicting precise temporal proposal boundaries and reliable action confidence in long and untrimmed real-world videos. In this paper, we propose an efficient and unified framework to generate temporal action proposals named Dense Boundary Generator (DBG), which draws inspiration from boundary-sensitive methods and implements boundary classification and action completeness regression for densely distributed proposals. In particular, the DBG consists of two modules: Temporal boundary classification (TBC) and Action-aware completeness regression (ACR). The TBC aims to provide two temporal boundary confidence maps by low-level two-stream features, while the ACR is designed to generate an action completeness score map by high-level action-aware features. Moreover, we introduce a dual stream BaseNet (DSB) to encode RGB and optical flow information, which helps to capture discriminative boundary and actionness features. Extensive experiments on popular benchmarks ActivityNet-1.3 and THUMOS14 demonstrate the superiority of DBG over the state-of-the-art proposal generator (e.g., MGG and BMN).

Recently, fully convolutional network (FCN) and dilated convolution have shown significantly improvment in semantic segmentation task. Deep residual network (ResNet) has shown strong ability in object recognition. However, FCN-based methods utilize intermediate layers and spatial context information ineffectively. Repeated 2-step striding in ResNet is harmful for segmentation tasks. In this paper, we propose a new semantic segmentation method based on FCN and ResNet. Here, we combine the dilated convolution designed for semantic segmentation with residual unit to enlarge receptive field of ResNet. Meanwhile, “pre-activation” method is used in dilated residual unit. We propose a new segmentation architecture which intergrate multiple intermediate layers and global context information. We extract low-level features of intermediate layers with multi-path refinement which consists of relu-conv unit and chained residual pooling. Global context is gathered by a pyramid pooling method which is connected to the final output of ResNet. Outputs of these 2 modules are then fused to reach high-resolution prediction. In this way, global and local information from pyramid pooling can be enhanced by multi-path refinement. Fully-connected conditional random field is added as a “post-processing” after fusion to receive accurate boundary performance. Our proposed approach achieves 74.7% IoU on Cityscapes benchmark.