Efficient crowd counting model using feature pyramid network and ResNeXt

Crowd counting is one of the most challenging issues in computer vision community for safety and security through surveillance systems. It has extensive range applications, such as disaster management, event detection, intelligence gathering analysis, public control, traffic monitoring, design spaces, anomaly detection military. Early approaches still encounter many like non-uniform density distribution, partial occlusion discrepancies scale perspective. To address above problems, feature pyramid networks are introduced deep convolution individuals crowd. The designed network extracted features at all resolutions constructed rapidly from only input image. This method achieves outperformance results compared to well-known on three standard crowd datasets.