{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/dual-path-multi-scale-fusion-networks-with","title":"Dual Path Multi-Scale Fusion Networks with Attention for Crowd Counting","arxiv_id":"1902.01115","date":"2019-02-04","proceeding":null,"authors":["Liang Zhu","Zhijian Zhao","Chao Lu","Yining Lin","Yao Peng","Tangren Yao"],"abstract":"The task of crowd counting in varying density scenes is an extremely\ndifficult challenge due to large scale variations. In this paper, we propose a\nnovel dual path multi-scale fusion network architecture with attention\nmechanism named SFANet that can perform accurate count estimation as well as\npresent high-resolution density maps for highly congested crowd scenes. The\nproposed SFANet contains two main components: a VGG backbone convolutional\nneural network (CNN) as the front-end feature map extractor and a dual path\nmulti-scale fusion networks as the back-end to generate density map. These dual\npath multi-scale fusion networks have the same structure, one path is\nresponsible for generating attention map by highlighting crowd regions in\nimages, the other path is responsible for fusing multi-scale features as well\nas attention map to generate the final high-quality high-resolution density\nmaps. SFANet can be easily trained in an end-to-end way by dual path joint\ntraining. We have evaluated our method on four crowd counting datasets\n(ShanghaiTech, UCF CC 50, UCSD and UCF-QRNF). The results demonstrate that with\nattention mechanism and multi-scale feature fusion, the proposed SFANet\nachieves the best performance on all these datasets and generates better\nquality density maps compared with other state-of-the-art approaches.","url_abs":"http://arxiv.org/abs/1902.01115v1","url_pdf":"http://arxiv.org/pdf/1902.01115v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"dual-path-multi-scale-fusion-networks-with","repo_url":"https://github.com/pxq0312/ASD-crowd-counting","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"dual-path-multi-scale-fusion-networks-with","repo_url":"https://github.com/pxq0312/SFANet-crowd-counting","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"crowd-counting","task_name":"Crowd Counting"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1902.01115","atlas_url":"https://app.syntology.ai/?focus=1902.01115","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}