Crowd counting with crowd attention convolutional neural network
Crowd counting is a challenging problem due to the scene complexity and scale variation. Although deep learning has achieved great improvement in crowd counting, scene complexity affects the judgement of these methods and they usually regard some objects as people mistakenly; causing potentially eno...
Сохранить в:
| Главные авторы: | , , |
|---|---|
| Формат: | Статья |
| Опубликовано: |
2022
|
| Предметы: | |
| Online-ссылка: | https://arxiv.org/abs/2204.07347 http://arxiv.org/abs/2204.07347 |
| Метки: |
Нет меток, Требуется 1-ая метка записи!
|
| _version_ | 1879173692181708800 |
|---|---|
| author | Chen, Jiwei Su, Wen Wang, Zengfu |
| author_browse | Chen, Jiwei Su, Wen Wang, Zengfu |
| author_facet | Chen, Jiwei Su, Wen Wang, Zengfu |
| author_sort | Chen, Jiwei |
| building | arXiv |
| collection | Acceso abierto |
| description | Crowd counting is a challenging problem due to the scene complexity and scale variation. Although deep learning has achieved great improvement in crowd counting, scene complexity affects the judgement of these methods and they usually regard some objects as people mistakenly; causing potentially enormous errors in the crowd counting result. To address the problem, we propose a novel end-to-end model called Crowd Attention Convolutional Neural Network (CAT-CNN). Our CAT-CNN can adaptively assess the importance of a human head at each pixel location by automatically encoding a confidence map. With the guidance of the confidence map, the position of human head in estimated density map gets more attention to encode the final density map, which can avoid enormous misjudgements effectively. The crowd count can be obtained by integrating the final density map. To encode a highly refined density map, the total crowd count of each image is classified in a designed classification task and we first explicitly map the prior of the population-level category to feature maps. To verify the efficiency of our proposed method, extensive experiments are conducted on three highly challenging datasets. Results establish the superiority of our method over many state-of-the-art methods. |
| doi_str_mv | 10.1016/j.neucom.2019.11.064 |
| first_indexed | 2026-10-16T00:01:17Z |
| format | Article |
| fullrecord | <oai_dc:dc xmlns:oai_dc="http://www.openarchives.org/OAI/2.0/oai_dc/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.openarchives.org/OAI/2.0/oai_dc/ https://www.openarchives.org/OAI/2.0/oai_dc.xsd"><identifier>arxiv-2204.07347</identifier><datestamp>2022-04-18</datestamp>
<dc:title>Crowd counting with crowd attention convolutional neural network</dc:title>
<dc:creator>Chen, Jiwei</dc:creator>
<dc:creator>Su, Wen</dc:creator>
<dc:creator>Wang, Zengfu</dc:creator>
<dc:subject>Computer Vision and Pattern Recognition</dc:subject>
<dc:subject>Machine Learning</dc:subject>
<dc:description>Crowd counting is a challenging problem due to the scene complexity and scale variation. Although deep learning has achieved great improvement in crowd counting, scene complexity affects the judgement of these methods and they usually regard some objects as people mistakenly; causing potentially enormous errors in the crowd counting result. To address the problem, we propose a novel end-to-end model called Crowd Attention Convolutional Neural Network (CAT-CNN). Our CAT-CNN can adaptively assess the importance of a human head at each pixel location by automatically encoding a confidence map. With the guidance of the confidence map, the position of human head in estimated density map gets more attention to encode the final density map, which can avoid enormous misjudgements effectively. The crowd count can be obtained by integrating the final density map. To encode a highly refined density map, the total crowd count of each image is classified in a designed classification task and we first explicitly map the prior of the population-level category to feature maps. To verify the efficiency of our proposed method, extensive experiments are conducted on three highly challenging datasets. Results establish the superiority of our method over many state-of-the-art methods. </dc:description>
<dc:description>Accepted by Neurocomputing</dc:description>
<dc:date>2022-04-15</dc:date>
<dc:type>text</dc:type>
<dc:identifier>https://arxiv.org/abs/2204.07347</dc:identifier>
<dc:identifier>http://arxiv.org/abs/2204.07347</dc:identifier>
<dc:identifier>Neurocomputing, 382, pages 210--220, (2020), Elsevier</dc:identifier>
<dc:identifier>doi:10.1016/j.neucom.2019.11.064</dc:identifier>
</oai_dc:dc> |
| id | arxiv-2204.07347 |
| institution | arXiv |
| last_indexed | 2026-10-16T00:01:17Z |
| publishDate | 2022 |
| publishDateRange | 2022 |
| publishDateSort | 2022 |
| record_format | oai_dc |
| spelling | arxiv-2204.073472022-04-18 Crowd counting with crowd attention convolutional neural network Chen, Jiwei Su, Wen Wang, Zengfu Computer Vision and Pattern Recognition Machine Learning Crowd counting is a challenging problem due to the scene complexity and scale variation. Although deep learning has achieved great improvement in crowd counting, scene complexity affects the judgement of these methods and they usually regard some objects as people mistakenly; causing potentially enormous errors in the crowd counting result. To address the problem, we propose a novel end-to-end model called Crowd Attention Convolutional Neural Network (CAT-CNN). Our CAT-CNN can adaptively assess the importance of a human head at each pixel location by automatically encoding a confidence map. With the guidance of the confidence map, the position of human head in estimated density map gets more attention to encode the final density map, which can avoid enormous misjudgements effectively. The crowd count can be obtained by integrating the final density map. To encode a highly refined density map, the total crowd count of each image is classified in a designed classification task and we first explicitly map the prior of the population-level category to feature maps. To verify the efficiency of our proposed method, extensive experiments are conducted on three highly challenging datasets. Results establish the superiority of our method over many state-of-the-art methods. Accepted by Neurocomputing 2022-04-15 text https://arxiv.org/abs/2204.07347 http://arxiv.org/abs/2204.07347 Neurocomputing, 382, pages 210--220, (2020), Elsevier doi:10.1016/j.neucom.2019.11.064 |
| spellingShingle | Computer Vision and Pattern Recognition Machine Learning Chen, Jiwei Su, Wen Wang, Zengfu Crowd counting with crowd attention convolutional neural network |
| title | Crowd counting with crowd attention convolutional neural network |
| title_full | Crowd counting with crowd attention convolutional neural network |
| title_fullStr | Crowd counting with crowd attention convolutional neural network |
| title_full_unstemmed | Crowd counting with crowd attention convolutional neural network |
| title_short | Crowd counting with crowd attention convolutional neural network |
| title_sort | crowd counting with crowd attention convolutional neural network |
| topic | Computer Vision and Pattern Recognition Machine Learning |
| url | https://arxiv.org/abs/2204.07347 http://arxiv.org/abs/2204.07347 |
| work_keys_str_mv | AT chenjiwei crowdcountingwithcrowdattentionconvolutionalneuralnetwork AT suwen crowdcountingwithcrowdattentionconvolutionalneuralnetwork AT wangzengfu crowdcountingwithcrowdattentionconvolutionalneuralnetwork |