MVFNet: Multi-View Fusion Network for Efficient Video Recognition

13 Dec 2020  ·  Wenhao Wu, Dongliang He, Tianwei Lin, Fu Li, Chuang Gan, Errui Ding ·

Conventionally, spatiotemporal modeling network and its complexity are the two most concentrated research topics in video action recognition. Existing state-of-the-art methods have achieved excellent accuracy regardless of the complexity meanwhile efficient spatiotemporal modeling solutions are slightly inferior in performance... In this paper, we attempt to acquire both efficiency and effectiveness simultaneously. First of all, besides traditionally treating H x W x T video frames as space-time signal (viewing from the Height-Width spatial plane), we propose to also model video from the other two Height-Time and Width-Time planes, to capture the dynamics of video thoroughly. Secondly, our model is designed based on 2D CNN backbones and model complexity is well kept in mind by design. Specifically, we introduce a novel multi-view fusion (MVF) module to exploit video dynamics using separable convolution for efficiency. It is a plug-and-play module and can be inserted into off-the-shelf 2D CNNs to form a simple yet effective model called MVFNet. Moreover, MVFNet can be thought of as a generalized video modeling framework and it can specialize to be existing methods such as C2D, SlowOnly, and TSM under different settings. Extensive experiments are conducted on popular benchmarks (i.e., Something-Something V1 & V2, Kinetics, UCF-101, and HMDB-51) to show its superiority. The proposed MVFNet can achieve state-of-the-art performance with 2D CNN's complexity. read more

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Action Classification Kinetics-400 MVFNet-ResNet101 (ensemble, ImageNet pretrained, RGB only) Vid acc@1 79.1 # 39
Vid acc@5 93.8 # 31
Action Recognition Something-Something V1 MVFNet-R50EN Top 1 Accuracy 54.0 # 14
Action Recognition Something-Something V2 MVFNet-ResNet50 (center crop, 8+16 ensemble, ImageNet pretrained, RGB only) Top-1 Accuracy 66.3 # 22

Methods