Huan Yang

Yuchong Sun, Hongwei Xue, Ruihua Song et al. · microsoft-research

Large-scale video-language pre-training has shown significant improvement in video-language understanding tasks. Previous studies of video-language pretraining mainly focus on short-form videos (i.e., within 30 seconds) and sentences, leaving long-form video-language pre-training rarely explored. Directly learning representation from long-form videos and language may benefit many long-form video-language understanding tasks. However, it is challenging due to the difficulty of modeling long-range relationships and the heavy computational burden caused by more frames. In this paper, we introduce a Long-Form VIdeo-LAnguage pre-training model (LF-VILA) and train it on a large-scale long-form video and paragraph dataset constructed from an existing public dataset. To effectively capture the rich temporal dynamics and to better align video and language in an efficient end-to-end manner, we introduce two novel designs in our LF-VILA model. We first propose a Multimodal Temporal Contrastive (MTC) loss to learn the temporal relation across different modalities by encouraging fine-grained alignment between long-form videos and paragraphs. Second, we propose a Hierarchical Temporal Window Attention (HTWA) mechanism to effectively capture long-range dependency while reducing computational cost in Transformer. We fine-tune the pre-trained LF-VILA model on seven downstream long-form video-language understanding tasks of paragraph-to-video retrieval and long-form video question-answering, and achieve new state-of-the-art performances. Specifically, our model achieves 16.1% relative improvement on ActivityNet paragraph-to-video retrieval task and 2.4% on How2QA task, respectively. We release our code, dataset, and pre-trained models at https://github.com/microsoft/XPretrain.

Shengming Yin, Chenfei Wu, Huan Yang et al. · microsoft-research, pku

In this paper, we propose NUWA-XL, a novel Diffusion over Diffusion architecture for eXtremely Long video generation. Most current work generates long videos segment by segment sequentially, which normally leads to the gap between training on short videos and inferring long videos, and the sequential generation is inefficient. Instead, our approach adopts a ``coarse-to-fine'' process, in which the video can be generated in parallel at the same granularity. A global diffusion model is applied to generate the keyframes across the entire time range, and then local diffusion models recursively fill in the content between nearby frames. This simple yet effective strategy allows us to directly train on long videos (3376 frames) to reduce the training-inference gap, and makes it possible to generate all segments in parallel. To evaluate our model, we build FlintstonesHD dataset, a new benchmark for long video generation. Experiments show that our model not only generates high-quality long videos with both global and local coherence, but also decreases the average inference time from 7.55min to 26s (by 94.26\%) at the same hardware setting when generating 1024 frames. The homepage link is \url{https://msra-nuwa.azurewebsites.net/}

Jun Liang, Songsen Yu, Huan Yang

ResNets and its variants play an important role in various fields of image recognition. This paper gives another variant of ResNets, a kind of cross-residual learning networks called C-ResNets, which has less computation and parameters than ResNets. C-ResNets increases the information interaction between modules by densifying jumpers and enriches the role of jumpers. In addition, some meticulous designs on jumpers and channels counts can further reduce the resource consumption of C-ResNets and increase its classification performance. In order to test the effectiveness of C-ResNets, we use the same hyperparameter settings as fine-tuned ResNets in the experiments. We test our C-ResNets on datasets MNIST, FashionMnist, CIFAR-10, CIFAR-100, CALTECH-101 and SVHN. Compared with fine-tuned ResNets, C-ResNets not only maintains the classification performance, but also enormously reduces the amount of calculations and parameters which greatly save the utilization rate of GPUs and GPU memory resources. Therefore, our C-ResNets is competitive and viable alternatives to ResNets in various scenarios. Code is available at https://github.com/liangjunhello/C-ResNet

Huan Yang, Baoyuan Wang, Stephen Lin et al.

With the growing popularity of short-form video sharing platforms such as \em{Instagram} and \em{Vine}, there has been an increasing need for techniques that automatically extract highlights from video. Whereas prior works have approached this problem with heuristic rules or supervised learning, we present an unsupervised learning approach that takes advantage of the abundance of user-edited videos on social media websites such as YouTube. Based on the idea that the most significant sub-events within a video class are commonly present among edited videos while less interesting ones appear less frequently, we identify the significant sub-events via a robust recurrent auto-encoder trained on a collection of user-edited videos queried for each particular class of interest. The auto-encoder is trained using a proposed shrinking exponential loss function that makes it robust to noise in the web-crawled training data, and is configured with bidirectional long short term memory (LSTM)~\cite{LSTM:97} cells to better model the temporal structure of highlight segments. Different from supervised techniques, our method can infer highlights using only a set of downloaded edited videos, without also needing their pre-edited counterparts which are rarely available online. Extensive experiments indicate the promise of our proposed solution in this challenging unsupervised settin