Deep Feature Flow for Video Recognition
This addresses the challenge of making state-of-the-art image recognition networks practical for video applications, though it is incremental as it builds on existing flow-based methods.
The paper tackles the problem of slow per-frame evaluation in video recognition by introducing deep feature flow, which runs convolutional sub-networks only on sparse key frames and propagates features to other frames via flow fields, achieving significant speedup and improved accuracy validated on large-scale datasets.
Deep convolutional neutral networks have achieved great success on image recognition tasks. Yet, it is non-trivial to transfer the state-of-the-art image recognition networks to videos as per-frame evaluation is too slow and unaffordable. We present deep feature flow, a fast and accurate framework for video recognition. It runs the expensive convolutional sub-network only on sparse key frames and propagates their deep feature maps to other frames via a flow field. It achieves significant speedup as flow computation is relatively fast. The end-to-end training of the whole architecture significantly boosts the recognition accuracy. Deep feature flow is flexible and general. It is validated on two recent large scale video datasets. It makes a large step towards practical video recognition.