Created by Tingwu Wang at Sense Time, Beijing and also with Tsinghua University.
It is a course projects, and sorry we have only a linux port one.
This pedestrian detection is based on the work of DPM and libccv. If you want to cite the original work, please cite
@article{girshick15fastrcnn,
Author = {Ross Girshick},
Title = {Fast R-CNN},
Journal = {arXiv preprint arXiv:1504.08083},
Year = {2015}
}
@misc{libccv,
author = "Liu Liu",
title = "C-based/Cached/Core Computer Vision Library, A Modern Computer Vision Library",
howpublished = "https://github.com/liuliu/ccv"
}
This pedestrian detector is under the MIT License (refer to the LICENSE file for details).
- The libccv dependency.
-lm -lcblas -latlas -lpng -ljpeg -lfftw3 -lfftw3f
-lpthread -llinear -lavcodec -lavformat -lswscale
-lgsl -lgslcblas -lcuda -lcudart -lcublas -L/usr/local/cuda/lib64
The cuda, cblas, atlas is what you really need to get. Other packages, you either already have it in Ubuntu, or could easily install using a sudo trick. 2. OpenCV library with 2.6+ 3. g++ with 4.6 or earlier
A GPU that support CUDA-7, or CUDA-6, earlier versions are your own adventures, cause I don't know what will happen.
The data is a big issue. Check out the $PEDESTRIAN/data/whats in the data directory.png It lists all the data I used to run the projects. Take a look in the source file $PEDESTRIAN/pedestrian/, download what is necessary.
- make sure you download all the lib you need.
- make the ccv library by
cd lib
./configure && make
- download the data you need, check out the
pedestrian-detection-thu2015/data/whats in the data directory.png
- Using the g++ to compile the source code.
cd pedestrian-detection-thu2015/pedestrian/
g++ ...
Remember to add the ccv dependency in /lib/.dep
and the ccv lib you just make in the step 2
and the OpenCV lib of course. 5. The results is stored in the /data/results (or change the path if you want)