Name
Abstract | ||
|---|---|---|
With the rapid growth of population all over the world, crowd analysis has become a vital way to maintain public safety in crowded scenes like outdoor sports events. In this paper, we propose a method that can accurately estimate number of people and their distribution in a crowded scene. Inspired by current state-of-the-art methods, we use multi-column CNN with different reception fields as basic regressor. Also, VGG-16 pre-trained on ImageNet is used to generate deep features. These two parts were then merged together for the final 1 x 1 convolutional layer and thus density maps are generated. Since all layers in our model is convolutional layers, the input image can be of any size, and the model is easy to implement and train. After experimenting on several major crowd counting datasets, our method turns out to have higher accuracy comparing to other existing methods. |
Year | DOI | Venue |
|---|---|---|
2018 | 10.1007/978-3-030-00764-5_49 | ADVANCES IN MULTIMEDIA INFORMATION PROCESSING, PT III |
Keywords | Field | DocType |
Crowd counting,Density estimation,Crowd analysis | Density estimation,Computer vision,Population,Pattern recognition,Convolutional neural network,Computer science,Artificial intelligence,Crowd counting,Crowd analysis | Conference |
Volume | ISSN | Citations |
11166 | 0302-9743 | 0 |
PageRank | References | Authors |
0.34 | 13 | 5 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Zhiyuan Xue | 1 | 0 | 0.34 |
| Jie Shen | 2 | 0 | 1.69 |
| Xin Xiong | 3 | 0 | 2.70 |
| Chong Yuan | 4 | 0 | 0.34 |
| Yinlong Bian | 5 | 0 | 0.34 |