1504.ICCVPartial Person Re-Identification 论文笔记
2017-12-18 23:01
1076 查看
1504.ICCVPartial Person Re-Identification 论文笔记
作者针对之前一些模型都是假设给以完整行人检测图像probe情况下,进行行人重识别任务,独辟蹊径,主要解决reid最难处理的遮挡问题,研究给以不完整的行人图像块来进行ReID任务建模。
主要贡献是:
将应用在partial人脸识别中MSTR(multi-task sparse representation无需对齐,用稀疏特征表达,即词典的方法表示特征)方法迁移过来,在此基础上,添加约束减少错误对齐和匹配图像块的数目,引入图像块相似性评分机制,得到AMC(Ambiguity-sensitive Matching Classifier)模型,in order to model explicitly the diversity and ambiguity of the occlusion patterns in partial person re-id in our new sparse representation classification formulation. based on patch-level local-to-local
matching
另外,考虑了全局性的部件匹配,利用上空间布局信息,减少错匹配,提出SWM(sliding window matching )模型,based on a global-to-local matching by a detection-based matching model
然后,作者对上述模型的分类结果得分做加权融合,并做实验设置加权参数,验证了融合模型AWC-SWM模型AWC和SWC相较于不融合能取得更好的表现。
由于是提出的partial Re-ID模型,为与之前模型作对比,作者制作自己的数据集,在CAVIVR与i-LIDS扩展数据集上裁剪partial图像块,即得P-CAVIAR and P-iLIDS数据集,并将设计的AMC-SWM融合模型与MSTR/RDC/PRSVM/L1-norm/LFDA/KISSME/LADF/SRC做了不同数据集下的比较,验证了结果有效性。
另外,作者说明了带核的LFDA模型相较于LFDA模型对partialReID任务不会产生有益的贡献
最后作者提出改进方向:将解决多尺度图像匹配问题,以得分匹配形式引入到融合模型中
作者针对之前一些模型都是假设给以完整行人检测图像probe情况下,进行行人重识别任务,独辟蹊径,主要解决reid最难处理的遮挡问题,研究给以不完整的行人图像块来进行ReID任务建模。
主要贡献是:
将应用在partial人脸识别中MSTR(multi-task sparse representation无需对齐,用稀疏特征表达,即词典的方法表示特征)方法迁移过来,在此基础上,添加约束减少错误对齐和匹配图像块的数目,引入图像块相似性评分机制,得到AMC(Ambiguity-sensitive Matching Classifier)模型,in order to model explicitly the diversity and ambiguity of the occlusion patterns in partial person re-id in our new sparse representation classification formulation. based on patch-level local-to-local
matching
另外,考虑了全局性的部件匹配,利用上空间布局信息,减少错匹配,提出SWM(sliding window matching )模型,based on a global-to-local matching by a detection-based matching model
然后,作者对上述模型的分类结果得分做加权融合,并做实验设置加权参数,验证了融合模型AWC-SWM模型AWC和SWC相较于不融合能取得更好的表现。
由于是提出的partial Re-ID模型,为与之前模型作对比,作者制作自己的数据集,在CAVIVR与i-LIDS扩展数据集上裁剪partial图像块,即得P-CAVIAR and P-iLIDS数据集,并将设计的AMC-SWM融合模型与MSTR/RDC/PRSVM/L1-norm/LFDA/KISSME/LADF/SRC做了不同数据集下的比较,验证了结果有效性。
另外,作者说明了带核的LFDA模型相较于LFDA模型对partialReID任务不会产生有益的贡献
最后作者提出改进方向:将解决多尺度图像匹配问题,以得分匹配形式引入到融合模型中
相关文章推荐
- 【论文笔记】Person Transfer GAN to Bridge Domain Gap for Person Re-Identification
- 论文笔记之:Deep Attributes Driven Multi-Camera Person Re-identification
- 【论文笔记】In Defense of the Triplet Loss for Person Re-Identification
- 1505.Deep Ranking for Person Re-identification via Joint Representation Learning论文笔记
- 1606.End--to-End Comparative Attention Networks for Person Re-identification 论文笔记
- 1703.In Defense of the Triplet Loss for Person Re-Identification 论文阅读笔记
- part-aligned系列论文:1707.ICCV.Pedestrian Alignment Network for Large-scale Person Re-identification 论文
- 论文笔记:Improving Deep Visual Representation for Person Re-identification by Global and Local Image-language Association
- 【论文笔记】Margin Sample Mining Loss: A Deep Learning Based Method for Person Re-identification
- 1701.Re-ranking Person Re-identification with k-reciprocal Encoding--PRW 论文阅读笔记
- part-aligned系列论文:1707.Deeply-Learned Part-Aligned Representations for Person Re-Identification 论文笔记
- 论文笔记之---Joint Detection and Identification Feature Learning for Person Search
- [论文翻译]Pedestrian Alignment Network for Large-scale Person Re-Identification
- Re-ID:AlignedReID: Surpassing Human-Level Performance in Person Re-Identification 论文解析
- 论文导读(person re-identification)——Person Re-identification based on nonlinear ranking with difference
- 读论文系列:Deep transfer learning person re-identification
- Scalable Person Re-id:A Benchmark论文笔记
- 论文笔记(一)Deep Ranking for Person Re-Identification via Joint Representation Learning
- 论文阅读《Joint Learning of Single-image Cross-image Representations for Person Re-identification》