Real-time Local Feature with Global Visual Information Enhancement
This work addresses computational and robustness issues in local feature extraction for visual tasks, representing an incremental improvement.
The paper tackled the limited receptive field and computational inefficiency of deep learning-based local feature algorithms by introducing a global enhancement module and optimizing with deep reinforcement learning, achieving robust performance against visual interference and real-time operation on public benchmarks.
Local feature provides compact and invariant image representation for various visual tasks. Current deep learning-based local feature algorithms always utilize convolution neural network (CNN) architecture with limited receptive field. Besides, even with high-performance GPU devices, the computational efficiency of local features cannot be satisfactory. In this paper, we tackle such problems by proposing a CNN-based local feature algorithm. The proposed method introduces a global enhancement module to fuse global visual clues in a light-weight network, and then optimizes the network by novel deep reinforcement learning scheme from the perspective of local feature matching task. Experiments on the public benchmarks demonstrate that the proposal can achieve considerable robustness against visual interference and meanwhile run in real time.