FAR: Fourier Aerial Video Recognition
This addresses activity recognition in UAV videos, which is important for surveillance and monitoring applications, and represents a novel method rather than an incremental improvement.
The paper tackles UAV video activity recognition by developing Fourier Activity Recognition (FAR), which uses Fourier object disentanglement to separate small human agents from backgrounds and Fourier Attention to model long-range dependencies, achieving relative improvements of 8.02% to 38.69% in top-1 accuracy and up to 3 times faster performance over prior works.
We present an algorithm, Fourier Activity Recognition (FAR), for UAV video activity recognition. Our formulation uses a novel Fourier object disentanglement method to innately separate out the human agent (which is typically small) from the background. Our disentanglement technique operates in the frequency domain to characterize the extent of temporal change of spatial pixels, and exploits convolution-multiplication properties of Fourier transform to map this representation to the corresponding object-background entangled features obtained from the network. To encapsulate contextual information and long-range space-time dependencies, we present a novel Fourier Attention algorithm, which emulates the benefits of self-attention by modeling the weighted outer product in the frequency domain. Our Fourier attention formulation uses much fewer computations than self-attention. We have evaluated our approach on multiple UAV datasets including UAV Human RGB, UAV Human Night, Drone Action, and NEC Drone. We demonstrate a relative improvement of 8.02% - 38.69% in top-1 accuracy and up to 3 times faster over prior works.