Efficient and Real-Time Particle Detection via Encoder-Decoder Network

Particle detection aims to locate and count valid particles in pad images accurately. However, existing methods fail to achieve both high detection accuracy and inference efficiency in real applications. In order to keep a good trade-off between inference efficiency and accuracy, we propose a computation-efficient particle detection network (PAD-Net) with an encoder-decoder architecture. For the encoder part, MobileNetV3 is tailored to greatly reduce parameters at a little cost of accuracy drop. And the decoder part is designed based on the lightweight RefineNet, which further boosts particle detection performance. Besides, the proposed network is equipped with the adaptive attention loss (termed AAL), which improves the detection accuracy with a negligible increase in computation cost. Finally, we employ a knowledge distillation strategy to further boost the final detection performance of PAD-Net without increasing its parameters and floating-point operations (FLOPs). Experimental results on the real datasets demonstrate that our methods can achieve high-accuracy and real-time detection performance on valid particles compared with the state-of-the-art methods.