traffic_video_analysis

This work is the official codes for paper: S. Zhang, G. Wu, J. Costeira, J. Moura, “Deep Understanding of Traffic Density from Large-Scale Web Camera Data”, accepted by Conference on Computer Vision and Pattern Recognition (CVPR) 2017.

Please email the author [email protected] if you are interested in the training code.

Please cite the following paper if you found the code is helpful for you:

S. Zhang, G. Wu, J. Costeira, J. Moura, “Understanding Traffic Density from Large-Scale Web Camera Data”, accepted by Conference on Computer Vision and Pattern Recognition (CVPR) 2017.

@article{szhangwebCam, title={Understanding Traffic Density from Large-Scale Web Camera Data}, author={Gon{S. Zhang, G. Wu, J. Costeira, J. Moura}, journal={arXiv preprint arXiv:1703.05868}, year={2017} }

Usage:

caffe_script:

extract_hypercolumn.py: The caffe script for extracting hypercolumn features. It uses the resnet-152 and combine res2a, res3a and res4a feature into a feature volumn called hypercolumn. Please set the root of caffe inside the script. It will read the images stored in the ../file_list/image_name_list.txt. Please feel free to change it to your file location. It assume the image format is jpg and save the hypercolumn feature with the same name but with extention ".resnet_hypercolumn"