Geometric Lp-norm Feature Pooling for Image Classification
2011-09-05 16:15
453 查看
by Jiashi Feng, Bingbing Ni, Qi Tian and Shuicheng Yan.
Average pooling: 统计每个visual word在图像中的出现次数,假设visual work $k$ 在一幅图像位置 $m$ 响应记为 $v_m^k$,则所有 $M$ 个图像位置的响应为一个 $M$ 的向量 $\textbf{v}^k$
\[f_a(\textbf{v}^k)=\frac{1}{M}\sum_{m=1}^M v_m^k\]
本文方法基于每个visual word的位置分布能够反映出类别信息的假设,在pooling的时候考虑了visual word的位置信息
\[f_g(\textbf{v}^k;\textbf{w}^k)=\sum_{m=1}^Mw_m^kv^k\]
对于每个类别,每个visual word都有一个位置权重map $\textbf{w}^k$,这是作为模型参数有one-vs-all方法学出。当 $\textbf{w}^k$ 所有元素都相等时,则退化为average pooling
假设:For images from a specific class, their visual features indexed by the same visual word often share similar spatial distribution.
Average pooling: 统计每个visual word在图像中的出现次数,假设visual work $k$ 在一幅图像位置 $m$ 响应记为 $v_m^k$,则所有 $M$ 个图像位置的响应为一个 $M$ 的向量 $\textbf{v}^k$
\[f_a(\textbf{v}^k)=\frac{1}{M}\sum_{m=1}^M v_m^k\]
本文方法基于每个visual word的位置分布能够反映出类别信息的假设,在pooling的时候考虑了visual word的位置信息
\[f_g(\textbf{v}^k;\textbf{w}^k)=\sum_{m=1}^Mw_m^kv^k\]
对于每个类别,每个visual word都有一个位置权重map $\textbf{w}^k$,这是作为模型参数有one-vs-all方法学出。当 $\textbf{w}^k$ 所有元素都相等时,则退化为average pooling
假设:For images from a specific class, their visual features indexed by the same visual word often share similar spatial distribution.
相关文章推荐
- Task-Driven Feature Pooling for Image Classification(一种监督的pooling方法——TDFpooling)
- 图像分类之特征学习ECCV-2010 Tutorial: Feature Learning for Image Classification
- 论文笔记(一)Re-ranking by Multi-feature Fusion with Diffusion for Image Retrieval
- 《PCANet: A Simple Deep Learning Baseline for Image Classification》精读笔记
- 结合人类视觉注意力 - Residual Attention Network for Image Classification
- 《PCANet: A Simple Deep Learning Baseline for Image Classification》中文翻译总结
- HD-CNN: HIERARCHICAL DEEP CONVOLUTIONAL NEURAL NETWORK FOR IMAGE CLASSIFICATION(泛读)
- [Paper 学习笔记]PCANet: A Simple Deep Learning Baseline for Image Classification?
- 读后感 Spectral–Spatial Residual Network for Hyperspectral Image Classification: A 3-D Deep Learning Fr
- Linear Spatial Pyramid Matching using Sparse Coding for Image Classification (阅读)
- Locality-constrained Linear Coding for Image Classification(阅读)
- Feature Selection for Classification: A Review(1)
- 吴恩达 深度学习 编程作业(1-4)- Building your DNN & DNN for Image Classification: Application
- Assignment | 01-week4 -Deep Neural Network for Image Classification: Application
- Multiple Kernels for Image Classification
- 论文PCANet: A Simple Deep Learning Baseline for Image Classification?的matlab源码解读(四)
- 论文笔记:Histopathological Image Classification Using Discriminative Feature-Oriented Dictionary Learnin
- Trained image classification models for Keras
- Feature Selection for Classification: A Review(二)
- Fisher vector for image classification