Scale Aware Dynamic Crowd Monitoring in Disaster Management

Frequent natural disasters pose significant challenges for emergency responders in managing post-disaster scenarios. Assessing the number of people affected is crucial for effective resource allocation and planning. Traditional crowd counting methods, like manual counts or ground-based photography, are often impractical in disaster zones. Unmanned aerial vehicles (UAVs), or drones, offer a promising solution due to their ability to capture high-resolution images from a bird's-eye view. Proposed Model is based on multi scale feature extraction from the input image and this model contains base as VGG-16 backbone and the model is build up of 3 parallel branches each responsible for specific purpose such as scale awareness, crowd distribution in a region and multi scale (low and high resolution) feature extraction. This attention based network help in generating the high resolution density map. The proposed model is tested on Shanghai tech A and B dataset, outperforms various state-of-art approaches.