A MPI programming model for fast bird nest detection on the railway catenary

Abstract

We propose a MIP programming model for the bird’s nest detection on the railway catenary, which performs coarse-to-fine strategy based on a cascaded YOLO network, and calculates the coarse-level and fine-level detection in parallel for different detected images. Due to the optimization of the parallel pipeline acceleration model, the deep learning network has a running speed equivalent to that of the single-stage network, which can perform real-time detection of bird’s nest.