Memory Warps for Learning Long-Term Online Video Representations
This work addresses the need for real-time video analysis in applications like surveillance or autonomous driving by improving efficiency and accuracy in online object detection, though it is incremental in its approach to memory and warping techniques.
The paper tackles the problem of learning efficient and accurate online video representations for object detection by proposing a memory-based method that warps features over time to compensate for motion, achieving a 2.3 times speed-up with only a 0.9% mAP drop on ImageNet-VID compared to prior offline methods.
This paper proposes a novel memory-based online video representation that is efficient, accurate and predictive. This is in contrast to prior works that often rely on computationally heavy 3D convolutions, ignore actual motion when aligning features over time, or operate in an off-line mode to utilize future frames. In particular, our memory (i) holds the feature representation, (ii) is spatially warped over time to compensate for observer and scene motions, (iii) can carry long-term information, and (iv) enables predicting feature representations in future frames. By exploring a variant that operates at multiple temporal scales, we efficiently learn across even longer time horizons. We apply our online framework to object detection in videos, obtaining a large 2.3 times speed-up and losing only 0.9% mAP on ImageNet-VID dataset, compared to prior works that even use future frames. Finally, we demonstrate the predictive property of our representation in two novel detection setups, where features are propagated over time to (i) significantly enhance a real-time detector by more than 10% mAP in a multi-threaded online setup and to (ii) anticipate objects in future frames.