CVPR2017: Learning Deep Context-aware Features over Body and Latent Parts for
2017-09-12 17:14
525 查看
论文查看
作者是中科院的Dangwei Li等。这篇工作是multi-class person identification tasks,主要创新有三点:(1)用空洞卷积(dilated conv)进行多尺度特征提取,减少传统CNN提取特征的信息损失;(2)利用Spatial Transformer Networks (STN,其中作者设置了三个参数限制) 提取可变的body-part,相比较于rigid divid, 能减少背景的影响; (3)将full body特征和parts特征融合,在identification classification 指导下,学习网络参数。
![](https://oscdn.geek-share.com/Uploads/Images/Content/201709/12/d09fe65241cde9498239a6d73d5280b1)
Yu, Fisher, and Vladlen Koltun. “Multi-scale context aggregation by dilated convolutions.” arXiv preprint arXiv:1511.07122 (2015).
Datasets: 第一段先整体介绍下哪些数据集上进行实验,其他段再分数据集单独介绍。
Protocols:介绍哪个数据集上进行哪些评估。
Model: 指出除了总的网络外,为了单独分析full body and body parts,抽出了两个sub models.
Optimization: 在caffe上实现,BP计算梯度,学习率等。
Data preprocessing :160x60, 1.0/256, image horizontally reflect
For the CUHK03 dataset, we compare our method with many existing
approaches, including XX, XX…..
Compared with XXXX, such as XXX, the proposed XXXX improves the Rank-1 identification rate by 11.66% and 13.29% on the labeled and detected datasets respectively.
Compared with XXXXX, our XXXX improves the Rank-1 identification rate by 2.93% and mAP by 4.22%.
选择一个数据集Market1501进行评估,分别设定dilated ratio 为1,2,3,4,指出3是个合适的选择(4的时候比3进展了一点点)
(2)可视化的方式展示学习到的部分和硬性分割的部分区别:学习到的部分由较大的重叠,同时包含较少的背景
(3)Effectiveness of location loss 评估约束的作用。
在基于部分的网络上评估约束(Market1501)
————————————————————————————————
关于dilated conv, 摘自知乎:
作者:谭旭
链接:https://www.zhihu.com/question/54149221/answer/192025860
来源:知乎
著作权归作者所有。
![](https://oscdn.geek-share.com/Uploads/Images/Content/201709/12/9e68a869786d49c1f7f3bf056e49e5e9)
作者是中科院的Dangwei Li等。这篇工作是multi-class person identification tasks,主要创新有三点:(1)用空洞卷积(dilated conv)进行多尺度特征提取,减少传统CNN提取特征的信息损失;(2)利用Spatial Transformer Networks (STN,其中作者设置了三个参数限制) 提取可变的body-part,相比较于rigid divid, 能减少背景的影响; (3)将full body特征和parts特征融合,在identification classification 指导下,学习网络参数。
Yu, Fisher, and Vladlen Koltun. “Multi-scale context aggregation by dilated convolutions.” arXiv preprint arXiv:1511.07122 (2015).
Part 4 Experiments
4.1 Datasets and protocols
本节分成Datasets和protocols两部分。Datasets: 第一段先整体介绍下哪些数据集上进行实验,其他段再分数据集单独介绍。
Protocols:介绍哪个数据集上进行哪些评估。
4.2 Implementation Details
分为Model , optimization 和 Data preprocessing 三部分Model: 指出除了总的网络外,为了单独分析full body and body parts,抽出了两个sub models.
Optimization: 在caffe上实现,BP计算梯度,学习率等。
Data preprocessing :160x60, 1.0/256, image horizontally reflect
4.3 Comparison with state-of-the-art methods
分数据集进行比较,每个数据集一个单独的部分。For the CUHK03 dataset, we compare our method with many existing
approaches, including XX, XX…..
Compared with XXXX, such as XXX, the proposed XXXX improves the Rank-1 identification rate by 11.66% and 13.29% on the labeled and detected datasets respectively.
Compared with XXXXX, our XXXX improves the Rank-1 identification rate by 2.93% and mAP by 4.22%.
4.4 Effectiveness of MSCAN(多尺度网络)
To determine the effectiveness of …., we explore four variants of … to learn IDE feature based on the whole body image.选择一个数据集Market1501进行评估,分别设定dilated ratio 为1,2,3,4,指出3是个合适的选择(4的时候比3进展了一点点)
4.5 Effectiveness of Latent part location
(1)Learned parts vs. rigid parts:选择Market1501比较学习到部分和硬性指定的部分对结果的影响。(2)可视化的方式展示学习到的部分和硬性分割的部分区别:学习到的部分由较大的重叠,同时包含较少的背景
(3)Effectiveness of location loss 评估约束的作用。
在基于部分的网络上评估约束(Market1501)
4.6 Cross-dataset Evaluation
在其他数据集上的模型迁移到VIPeR,分别验证了直接迁移和微调后的性能。————————————————————————————————
关于dilated conv, 摘自知乎:
作者:谭旭
链接:https://www.zhihu.com/question/54149221/answer/192025860
来源:知乎
著作权归作者所有。
相关文章推荐
- 论文阅读:End-to-End Learning of Deformable Mixture of Parts and Deep Convolutional Neural Networks for H
- End-to-End Learning of Deformable Mixture of Parts and Deep Convolutional Neural Networks for Human
- 立体匹配——End-to-End Learning of Geometry and Context for Deep Stereo Regression
- End-to-End Learning of Geometry and Context for Deep Stereo Regression
- 行人属性“Weakly-supervised Learning of Mid-level Features for Pedestrian Attribute Recognition and Loca”
- 1607.CVPR-Joint Learning of Single-image and Cross-image Representations for Person ReID 论文笔记
- 台大李宏毅Machine Learning 2017Fall学习笔记 (10)Tips for Deep Learning
- PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation
- [Stereo_cnn][cvpr16]Efficient Deep Learning for Stereo Matching(未完成)
- Learning Deep Features for Discriminative Localization
- kernel learning for extrinsic classification of manifold features+CVPR13
- Deep Learning Strong Parts for Pedestrian Detection
- Deep Direct Reinforcement Learning for Financial Signal Representation and Trading
- PredNet --- Deep Predictive coding networks for video prediction and unsupervised learning --- 论文笔记
- Deep learning for CWD, POS and NER
- Learning Deep Features for Discriminative Localization论文笔记以及Caffe实现
- 配置和运行 MatchNet CVPR 2015 MatchNet: Unifying Feature and Metric Learning for Patch-Based Matching
- 论文阅读-《Learning Deep Features for Discriminative Localization》
- 论文阅读-《Learning Deep Features for Discriminative Localization》
- 【论文导读】Scalable and Accurate Deep Learning for Electronic Health Records