
trecvid_inf.mat contains informations for compiling a-TRECVID dataset

trecvid_inf = 

         att_name: {500x1 cell}            500 attributes names
             path: {266473x1 cell}         path of the 266,473 images
    query_att_idx: [1x126 double]          index of 126 attributes used
       video_name: {1x11640 cell}          name of videos
      video_label: [1x266473 double]       video labels for the 266,473 images
     train_idx_2000: [1x2000 double]			 index of 2000 images for training
     train_idx_4000: [1x4000 double]	     index of 4000 images for training
     train_idx_6000: [1x6000 double]		   index of 6000 images for training
     train_idx_8000: [1x8000 double]       index of 8000 images for training
    train_idx_10000: [1x10000 double]      index of 10000 images for training

********************************************************
You will need TRECVID 2011 SIN dataset in order to get a-TRECVID: http://www-nlpir.nist.gov/projects/tv2011/tv2011.html#sin. The dataset is available to all the participants of TRECVID SIN task. However it cannot be redistributed outside the participants' groups. Please email yuxinnan@ee.columbia.edu for the data.
You will also need to download the TRECVID 2011 collaborative annotation http://mrim.imag.fr/tvca/ in order to get the attribute labels. 
Please email yuxinnan@ee.columbia.edu if you have encountered any problems. 

********************************************************
We used the following weak attribtues for experiments:
1. 126 query attribute classifier scores.
	trained with liblinear with C = 1 http://www.csie.ntu.edu.tw/~cjlin/liblinear/
2. 2,659 classemes scores
	http://www.cs.dartmouth.edu/~lorenzo/projects/classemes/
3. 2,000 discriminative attributes
	As described in A. Farhadi, I. Endres, D. Hoiem, and D.A. Forsyth, Describing Objects by their Attributes, CVPR 2009, and our CVPR2012 paper
4. 1,215 random projection and latent variables
	Refer to our CVPR2012 paper.

*********************************************************
Please cite the following papers when using the data/code:
Felix X. Yu; Rongrong Ji; Ming-Hen Tsai; Guangnan Ye; Shih-Fu Chang.
Weak attributes for large-scale image retrieval, CVPR 2012

Please also consider to cite our technical report covering more experiments:
Felix X. Yu; Rongrong Ji; Ming-Hen Tsai; Guangnan Ye; Shih-Fu Chang.
Experiments of image retrieval using weak attributes, Technical Report # CUCS 005-12

*************************************************
Report questions to Felix X. Yu: yuxinnan@ee.columbia.edu
