- Pedestrian Detection
This code produces the results presented in http://arxiv.org/abs/1409.5209 on the Caltech dataset (optical flow used; so it doesn't work on the INRIA dataset) with BING as the pre-processor.
If you use this code in your research, please cite our papers:
@inproceedings{PaisitkriangkraiSH14a,
author = {Sakrapee Paisitkriangkrai and
Chunhua Shen and
Anton {van den Hengel}},
title = {Strengthening the Effectiveness of Pedestrian Detection with Spatially Pooled Features},
booktitle = {Proc. European Conf. Comp. Vis.},
year = {2014},
ee = {http://arxiv.org/abs/1407.0786},
}
@inproceedings{PaisitkriangkraiSH14b,
author = {Sakrapee Paisitkriangkrai and
Chunhua Shen and
Anton van den Hengel},
title = {Pedestrian Detection with Spatially Pooled Features and Structured Ensemble Learning},
journal = {CoRR},
year = {2014},
ee = {http://arxiv.org/abs/1409.5209},
}
-
The current demo contains a few test images from Caltech Pedestrian data sets (set07, V004).
-
(a) Compile optical flow source code if needed by (Precompiled files provided already! You may not need to compile your own version)
sh> ./mex_optical.sh
- (b) Run demo.m (This will generate the ROC curve on the Caltech dataset set07, V004. It will download the data first ~400M.)
matlab> demo
WARNING: It may take 2 to 4 hours to get the result, depending on your machine. You should see a plot as below.
- More Caltech Pedestrian test data can be obtained from
http://www.vision.caltech.edu/Image_Datasets/CaltechPedestrians/datasets/USA/
-
NB: set00-set05 are used for training and set06-set10 are used for testing see http://www.vision.caltech.edu/Image_Datasets/CaltechPedestrians/
-
The data set is in seq video format. Download MATLAB functions for read/write seq video files from http://vision.ucsd.edu/~pdollar/toolbox/doc/index.html
The code has been tested to run on Ubuntu 14.04LTS (kernel: Linux 3.13.0-39-generic #66-Ubuntu SMP x86_64 GNU/Linux), Matlab 2013a.
