We propose a real-time convolutional neural network (CNN) training and compression method for delivering high-quality live video even in a poor network environment. The server delivers a low-resolution video segment along with the corresponding CNN for super resolution (SR), after which the client applies the CNN to the segment in order to recover high-resolution video frames. To generate a trained CNN corresponding to a video segment in real-time, our method rapidly increases the training accuracy by promoting the overfitting property of the CNN while also using curriculum-based training. In addition, assuming that the pretrained CNN is already downloaded on the client side, we transfer only residual values between the updated and pretrained CNN parameters. These values can be quantized with low bits in real time while minimizing the amount of loss, as the distribution range is significantly narrower than that of the updated CNN. Quantitatively, our neural-enhanced adaptive live streaming pipeline (NEALS) achieves higher SR accuracy and a lower CNN compression loss rate within a constrained training time compared to the state-of-the-art CNN training and compression method. NEALS achieves 15 to 48% higher quality of the user experience compared to state-of-the-art neural-enhanced live streaming systems.

We present a new interactive playback method to enhance 360° viewing experiences. Our method automatically rotates the virtual camera of a 360° panoramic video (360° video) player during interactive viewing to guide the viewer through the most important regions of the video. With this method, the viewer can watch a 360° video with minimum efforts to find important events in a scene both in interactive (e.g., HMD) and less-interactive (e.g., PC and TV) viewing environments. To estimate the importance of each viewing direction, we combine spatial and temporal saliency with cluster-based weighting. A maximum backward cumulative importance volume (MBCIV) is then constructed by accumulating this importance in the video space. During playback, which uses a forward tracing scheme through the MBCIV, the initial optimal path is found based on the viewer’s viewing direction. A smooth path is then derived using penalized curve fitting. Finally, the virtual camera is rotated to follow the path. The experiments and user studies demonstrate that our method allows the viewer to effectively enjoy 360° videos with minimum interaction efforts, or even through a non-interactive display.

We present a hybrid approach that segments an object by using both color and depth information obtained from views captured from a low-cost RGBD camera and sparsely-located color cameras. Our system begins with generating dense depth information of each target image by using Structure from Motion and Joint Bilateral Upsampling. We formulate the multi-view object segmentation as the Markov Random Field energy optimization on the graph constructed from the superpixels. To ensure inter-view consistency of the segmentation results between color images that have too few color features, our local mapping method generates dense inter-view geometric correspondences by using the dense depth images. Finally, the pixel-based optimization step refines the boundaries of the results obtained from the superpixel-based binary segmentation. We evaluate the validity of our method under various capture conditions such as numbers of views, rotations, and distances between cameras. We compared our method with the state-of-the-art methods that use the standard multi-view datasets. The comparison verified that the proposed method works very efficiently especially in a sparse wide-baseline capture environment.