RPN_BF-RPN-pedestrian
所属分类:图形图像处理
开发工具:matlab
文件大小:5675KB
下载次数:40
上传日期:2017-04-23 15:46:52
上 传 者:
yxsky
说明: 行人检测.南开大学多媒体实验室的源码。是matlab。数据集是在caltech上。
(pedestrian detection)
文件列表:
RPN_BF-RPN-pedestrian (0, 2016-10-17)
RPN_BF-RPN-pedestrian\LICENSE (3761, 2016-10-17)
RPN_BF-RPN-pedestrian\datasets (0, 2016-10-17)
RPN_BF-RPN-pedestrian\datasets\caltech (0, 2016-10-17)
RPN_BF-RPN-pedestrian\datasets\caltech\extract_img_anno.m (562, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments (0, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\+Dataset (0, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\+Dataset\caltech_test.m (551, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\+Dataset\caltech_trainval.m (559, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\+Dataset\private (0, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\+Dataset\private\voc0712_devkit.m (75, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\+Dataset\private\voc2007_devkit.m (75, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\+Dataset\private\voc2012_devkit.m (75, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\+Faster_RCNN_Train (0, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\+Faster_RCNN_Train\do_generate_bf_proposal_caltech.m (3869, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\+Faster_RCNN_Train\do_proposal_test_caltech.m (4277, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\+Faster_RCNN_Train\do_proposal_train_caltech.m (859, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\+Faster_RCNN_Train\set_cache_folder_caltech.m (432, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\+Model (0, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\+Model\VGG16_for_rpn_pedestrian_caltech.m (1063, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\script_rpn_bf_pedestrian_VGG16_caltech.m (7992, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\script_rpn_bf_pedestrian_VGG16_caltech_demo.m (6139, 2016-10-17)
RPN_BF-RPN-pedestrian\experiments\script_rpn_pedestrian_VGG16_caltech.m (2944, 2016-10-17)
RPN_BF-RPN-pedestrian\external (0, 2016-10-17)
RPN_BF-RPN-pedestrian\external\caffe (0, 2016-10-17)
RPN_BF-RPN-pedestrian\external\caffe\matlab (0, 2016-10-17)
RPN_BF-RPN-pedestrian\external\caffe\matlab\caffe_faster_rcnn (0, 2016-10-17)
RPN_BF-RPN-pedestrian\external\caffe\matlab\caffe_faster_rcnn\+caffe (0, 2016-10-17)
RPN_BF-RPN-pedestrian\external\caffe\matlab\caffe_faster_rcnn\+caffe\+test (0, 2016-10-17)
RPN_BF-RPN-pedestrian\external\caffe\matlab\caffe_faster_rcnn\+caffe\+test\test_net.m (3971, 2016-10-17)
RPN_BF-RPN-pedestrian\external\caffe\matlab\caffe_faster_rcnn\+caffe\+test\test_solver.m (1422, 2016-10-17)
RPN_BF-RPN-pedestrian\external\caffe\matlab\caffe_faster_rcnn\+caffe\Blob.m (3046, 2016-10-17)
RPN_BF-RPN-pedestrian\external\caffe\matlab\caffe_faster_rcnn\+caffe\Layer.m (982, 2016-10-17)
RPN_BF-RPN-pedestrian\external\caffe\matlab\caffe_faster_rcnn\+caffe\Net.m (7437, 2016-10-17)
RPN_BF-RPN-pedestrian\external\caffe\matlab\caffe_faster_rcnn\+caffe\Solver.m (1901, 2016-10-17)
RPN_BF-RPN-pedestrian\external\caffe\matlab\caffe_faster_rcnn\+caffe\get_net.m (1207, 2016-10-17)
... ...
# Is Faster R-CNN Doing Well for Pedestrian Detection?
By Liliang Zhang, Liang Lin, Xiaodan Liang, Kaiming He
### Introduction
This code is relative to an [arXiv tech report](https://arxiv.org/abs/1607.07032), which is accepted on ECCV 2016.
The RPN code in this repo is written based on the MATLAB implementation of Faster R-CNN. Details about Faster R-CNN are in: [ShaoqingRen/faster_rcnn](https://github.com/ShaoqingRen/faster_rcnn).
This BF code in this repo is written based on Piotr's Image & Video Matlab Toolbox. Details about Piotr's Toolbox are in: [pdollar/toolbox](https://github.com/pdollar/toolbox).
This code has been tested on Ubuntu 14.04 with MATLAB 2014b and CUDA 7.5.
### Citing RPN+BF
If you find this repo useful in your research, please consider citing:
@article{zhang2016faster,
title={Is Faster R-CNN Doing Well for Pedestrian Detection?},
author={Zhang, Liliang and Lin, Liang and Liang, Xiaodan and He, Kaiming},
journal={arXiv preprint arXiv:1607.07032},
year={2016}
}
### Requirements
0. `Caffe` build for RPN+BF (see [here](https://github.com/zhangliliang/caffe/tree/RPN_BF))
- If the mex in 'external/caffe/matlab/caffe_faster_rcnn' could not run under your system, please follow the [instructions](https://github.com/zhangliliang/caffe/tree/RPN_BF) on our Caffe branch to compile and replace the mex.
0. MATLAB
0. GPU: Titan X, K40c, etc.
**WARNING**: The `caffe_.mexa***` in `external/caffe/matlab/caffe_faster_rcnn` might be not compatible with your computer. If so, please try to compile [this Caffe version](https://github.com/zhangliliang/caffe/tree/RPN_BF) and replace it.
### Testing Demo
0. Download `VGG16_caltech_final.zip` from [BaiduYun](https://pan.baidu.com/s/1miNdKZe),or [Onedrive](https://1drv.ms/u/s!AgVYvWT--3HKhBhVNhWaSNcV2U0-) and unzip it in the repo folder.
0. Start MATLAB from the repo folder.
0. Run `faster_rcnn_build`
0. Run `script_rpn_bf_pedestrian_VGG16_caltech_demo` to see the detection results on some images collected in Internet.
### Training on Caltech (RPN)
0. Download "Matlab evaluation/labeling code (3.2.1)" as `external/code3.2.1` by run `fetch_data/fetch_caltech_toolbox.m`
0. Download the annotations and videos in [Caltech Pedestrian Dataset](http://www.vision.caltech.edu/Image_Datasets/CaltechPedestrians/datasets/USA/) and put them in the proper folder follow the instruction in the [website](http://www.vision.caltech.edu/Image_Datasets/CaltechPedestrians/).
0. Download the VGG-16 pretrain model and the relative prototxt in `VGG16_caltech_pretrain.zip` from [BaiduYun](http://pan.baidu.com/s/1nvGYOVR) or [OneDrive](https://1drv.ms/u/s!AgVYvWT--3HKhCwAD2i_JvgIOPrR), and unzip it in the repo folder. The md5sum for `vgg16.caffemodel` should be `e54292186923567dc14f21dee292ae36`.
0. Start MATLAB from the repo folder, and run `extract_img_anno` for extracting images in JPEG format and annotations in TEXT format from the Caltech dataset.
0. Run `script_rpn_pedestrian_VGG16_caltech` to train and test the RPN model on Caltech. Wait about half day for training and testing.
0. Hopefully it would give the evaluation results around ~14% MR after running.
### Training on Caltech (RPN+BF)
0. Follow the instruction in "Training on Caltech (RPN)" for obtaining the RPN model.
0. Run `script_rpn_bf_pedestrian_VGG16_caltech` to train and test the BF model on Caltech. Wait about two or three days for training and testing.
0. Hopefully it would give the evaluation results around ~10% MR after running.
近期下载者:
相关文件:
收藏者: