PVANET----Deep but Lightweight Neural Networks for Real-time Object Detection论文记录
2016-10-27 18:12
756 查看
arxiv上放出的物体检测的文章,在Pascal voc数据集上排第二。源码也已放出(https://github.com/sanghoon/pva-faster-rcnn),又可以慢慢把玩了。这篇文章遵循faster-rcnn“CNN feature extraction + region proposal + RoI classification”的pipeline,重新设计了feature extraction的网络结构。"The devil is in details",文章利用很多的cnn tricks,详述了网络设计的细节。
C.ReLU: Earlier building blocks in feature generation
C.ReLU是在ICML2016一篇文章提出。文章发现,CNN的初期阶段,神经元的激活值正好相反。C.ReLU把卷积输出的通道数减半,将输出与其负向输出级联,在没有损失正确率的情况下,获得两倍的加速。
Inception: Remaining building blocks in feature generation
Inception是GoogleNet的重要组成模块,却还没用在检测任务上。Inception中的1x1卷积核不仅能够增加网络的非线性,同时能够保持前一层的感受野,因此对小物体的检测有很好的作用。文中还把原来5x5的卷积核换成两个3x3的卷积核,减少参数,增加网络非线性和模块感受野。
HyperNet: Concatenation of multi-scale intermediate outputs
HyperNet将不同卷积阶段的卷积层级联起来,对同时需要分类和定位的检测任务来说有很好的效果。
论文的级联为:
combines 1) the last layer and 2) two intermediate layers whose scales are 2x and 4x of the last
layer, respectively.
The pvanet architecture
Deep network training
文章用了residual connections 和batch normalization加速网络收敛。BN层加在ReLU层后面,学习率根据plateau detection自动调整。
RPN用了25个anchor(5 scales(3,6,9,16,25),5 aspect ratios(0.5,0.557,1.0,1.5,2.0))。最后的全连接层使用了简单的SVD分解,map有部分降低,检测速度加快。
result
C.ReLU: Earlier building blocks in feature generation
C.ReLU是在ICML2016一篇文章提出。文章发现,CNN的初期阶段,神经元的激活值正好相反。C.ReLU把卷积输出的通道数减半,将输出与其负向输出级联,在没有损失正确率的情况下,获得两倍的加速。
Inception: Remaining building blocks in feature generation
Inception是GoogleNet的重要组成模块,却还没用在检测任务上。Inception中的1x1卷积核不仅能够增加网络的非线性,同时能够保持前一层的感受野,因此对小物体的检测有很好的作用。文中还把原来5x5的卷积核换成两个3x3的卷积核,减少参数,增加网络非线性和模块感受野。
HyperNet: Concatenation of multi-scale intermediate outputs
HyperNet将不同卷积阶段的卷积层级联起来,对同时需要分类和定位的检测任务来说有很好的效果。
论文的级联为:
combines 1) the last layer and 2) two intermediate layers whose scales are 2x and 4x of the last
layer, respectively.
The pvanet architecture
Deep network training
文章用了residual connections 和batch normalization加速网络收敛。BN层加在ReLU层后面,学习率根据plateau detection自动调整。
RPN用了25个anchor(5 scales(3,6,9,16,25),5 aspect ratios(0.5,0.557,1.0,1.5,2.0))。最后的全连接层使用了简单的SVD分解,map有部分降低,检测速度加快。
result
相关文章推荐
- 论文笔记:PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection
- 论文阅读--PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection
- 多尺度R-CNN论文笔记(4): PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection
- PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection
- 目标检测--PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection
- PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection
- PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection - arxiv 2016.08
- 读书笔记PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection
- [Paper note] PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection
- 目标检测--PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection
- READING NOTE: PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection
- PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection
- tensorfolw配置过程中遇到的一些问题及其解决过程的记录(配置SqueezeDet: Unified, Small, Low Power Fully Convolutional Neural Networks for Real-Time Object Detection for Autonomous Driving)
- 论文笔记 《Deep Neural Networks for Object Detection》
- 【论文笔记】Deep Neural Networks for Object Detection
- Deep Neural Networks for Object Detection论文翻译
- 论文阅读:Deep Neural Networks for Object Detection
- 【深度学习论文笔记:Recognition】:Deep Neural Networks for Object Detection
- 论文笔记| Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
- Deep Neural Networks for Object Detection(基于DNN的对象检测)