Skip to content
The codes for CVPR-2018 paper "Crowd Counting with Deep Negative Correlation Learning"
Branch: master
Clone or download
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
cmake initial commit Jun 11, 2018
data initial commit Jun 11, 2018
docker initial commit Jun 11, 2018
docs initial commit Jun 11, 2018
examples/crowd Delete deploy2.prototxt~ Oct 23, 2018
include/caffe Add files via upload Sep 8, 2018
matlab initial commit Jun 11, 2018
models initial commit Jun 11, 2018
python initial commit Jun 11, 2018
scripts initial commit Jun 11, 2018
src Add files via upload Sep 8, 2018
tools
CMakeLists.txt initial commit Jun 11, 2018
CONTRIBUTING.md initial commit Jun 11, 2018
CONTRIBUTORS.md initial commit Jun 11, 2018
INSTALL.md initial commit Jun 11, 2018
LICENSE initial commit Jun 11, 2018
Makefile initial commit Jun 11, 2018
Makefile.config initial commit Jun 11, 2018
Makefile.config.example initial commit Jun 11, 2018
README.md Update README.md Oct 22, 2018
caffe.cloc initial commit Jun 11, 2018
dmap.jpeg initial commit Jun 11, 2018
dmap.jpg initial commit Jun 11, 2018
dmap.png initial commit Jun 11, 2018

README.md

Crowd-Counting-with-Deep-Negative-Correlation-Learning

The codes for CVPR-2018 paper "Crowd Counting with Deep Negative Correlation Learning" in http://openaccess.thecvf.com/content_cvpr_2018/papers/Shi_Crowd_Counting_With_CVPR_2018_paper.pdf

To run this codes, what you need to do is as follows:

  1. Compiling the Caffe codes. You should have installed Caffe correctly. Then you can clone or download our codes and make some changes in Makefile.config to compile correctly.

  2. Preparing your data. The codes for preparing crowd counting dataset can be found in ¨examples/crowd/shanghaiA/predata¨. If you would like to run your own tasks, you have to write the codes by yourself.

  3. Training. In ¨examples/crowd¨, you can find the network prototxt and solver prototxt, and you should make some changes according to your tasks. If you want to use different ¨K¨ (K stands for the number of base regresors in the ensemble, we found K=64 gives the best performance in our tasks.), the ¨examples/crowd/shanghaiA/create_prototxt.py¨ can help to generate your network prototxt easily.

  4. Testing. If you use MAE and MSE as your evaluation metrics, you can monitor the testing results in training. You just need to add ¨layer { name: "mae" type: "MAELoss" bottom: "avgscore" bottom: "label" top: "mae" include { phase: TEST } }¨ and ¨layer { name: "mse" type: "MSELoss" bottom: "avgscore" bottom: "label" top: "mse" include { phase: TEST } }¨ to your network prototxt.

  5. Please carefully tune learning rate, we found it had a great influence to the training results in our tasks.

If you use this codes, please kindly cite our paper:

@InProceedings{Shi_2018_CVPR, author = {Shi, Zenglin and Zhang, Le and Liu, Yun and Cao, Xiaofeng and Ye, Yangdong and Cheng, Ming-Ming and Zheng, Guoyan}, title = {Crowd Counting With Deep Negative Correlation Learning}, booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)}, month = {June}, year = {2018} }

Please feel free to contact us if you still have any questions. Zenglin Shi: iezlshi@gmail.com Le Zhang: zhang.le@adsc.com.sg

You can’t perform that action at this time.