This folder contains demo code for “What If We Do Not Have Multiple Videos of the Same Action? -- Video Action Localization Using Web Images, CVPR16”
The main file is demo_Code.m.
In order to use the code, you need to compute Action Proposals and CNN features etc. We have include pre-computed proposals for ‘Spatio-Temporal Object Detection Proposals, ECCV’14’. Please see comments in the code for more details.
The video frames should be in folder given by “Videos_frame_Path”.
This software requires the following resources, which are already integrated. We have changed the codes of ([2] and [3]); changed version are included. [1] A. Vedaldi and K. Lenc. Matconvnet – convolutional neural networks for matlab. 2015 [2] M. Cho, S. Kwak, C. Schmid, and J. Ponce. Unsupervised object discovery and localization in the wild: Part-based matching with bottom-up region proposals. In CVPR, 2015. [3] M. Schmidt. Graphical model structure learning with l1- regularization. In Ph.D. Thesis, 2010.
The modified [1] and [2] can downloaded from this webpage under Matlab code: http://crcv.ucf.edu/projects/videolocalization_images//
In case of any comments, please drop me an email at waqas5163@gmail.com.
If you found this code useful in your research, please cite the following paper:
@InProceedings{Sultani_2016_CVPR, author = {Sultani, Waqas and Shah, Mubarak}, title = {What If We Do Not Have Multiple Videos of the Same Action? -- Video Action Localization Using Web Images}, booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)}, month = {June}, year = {2016} }
Thank you!