Motion Robust High-Speed Light-Weighted Object Detection With Event Camera

In this work, we propose a motion robust and high-speed detection pipeline which better leverages the event data. First, design an stream representation called temporal active focus (TAF), efficiently utilizes spatial-temporal asynchronous stream, constructing tensors to object motions. Then, module bifurcated folding (BFM), encodes rich information in TAF tensor at input layer of detector. Following this, lightweight detector agile (AED) plus simple but effective data augmentation method, enhance accuracy reduce model's parameter. Experiments on two typical real-scene camera datasets show that our method is competitive terms accuracy, efficiency, number parameters. By classifying objects into multiple levels based optical flow density metric, further illustrated robustness for with different velocities relative camera. The codes trained models are available https://github.com/HarmoniaLeo/FRLW-EvD .