We proposed a Scale-Aware Crowd Counting Network that can efficiently deal with the task of crowd counting under clutter background and scale variations
Unlike multi-view multi-scale, we propose to solve the multi-view crowd counting task through 3D feature fusion with 3D scene-level density maps, instead of the 2D ground-plane ones
The key reason for this surge in interest is the demand of automated complex crowd scene understanding that appears in computer vision applications such as surveillance, traffic monitoring, etc
We show that the proposed method is able to recover a major percentage of the performance drop. Synthetic-to-real transfer settings: In this setting, the goal is to train on synthetic dataset, while adapting to real-world dataset
In Neural Architecture Search-Count we propose a counting-oriented NAS framework with specific search space, search strategy, and supervision method, what we use to develop our Automatic Multi-Scale Network
We have presented a novel attention scaling based counting network that exploits attention masks and scaling factors to correct density estimations in regions of different density levels
We provide a comprehensive overview and comparison of three major design modules for deep learning models in crowd counting, deep neural network design, loss function, and supervisory signal
We present a novel network, Scale Tree Network, which consistently addresses the challenges of drastic scale variations, density changes, and complex background
We propose a novel Dilated-Scale-Aware Category-Attention ConvNet, which achieves multi-class object counting simultaneously only based on point-level annotations
This paper proposes an effective crowd localization framework, Independent Instance Map, which outputs independent instance maps to localize each head in crowd scenes
This paper proposes a novel crowd counting approach based on pyramidal scale module and global context module, dubbed Pyramidal Scale and global Context Network