The result.zip contains the groundtruth data of ¡°direct ground truth labeling¡±.  
According to our paper, we provide a pre-selected query samples list to every annotator. 
The annotator can labels his/her own selected images set, 
which is similar to the query sample based on his/her own visual perception.  
Consequently, after such process, for each query sample, 
there is a corresponding image list in which the image is similar to the query 
based on human perception and thus the list could be considered as the retrieval groundtruth. 
7 college students evolve in this annotation work. 
The result is build based on the average of 7 students¡¯ annotation results.

So, you can get the annotation result as the following:
1. Download the file result.zip
2. Unzip the file to get resultnew.txt

The format of this file is the following:

The location of Query image 1 at Corel Image Gallery 1 Million
The Number of The result Images for Query 1
The location of Result Image 1
Similarity to the Query([0,1])
The location of Result Image 2
Similarity to the Query([0,1])
¡­
The location of Query image 2
The Number of The result Images for Query 2
The location of Result Image 1
Similarity to the Query([0,1])
The location of Result Image 2
Similarity to the Query([0,1])
¡­

For example, 
Disk08/Photos/Mosaic/639005.JPG ( Query Image¡¯s location )
146 ( number of its ground truth result )
Disk08/Photos/Mosaic/639059.JPG ( Location of the 1st result )
1.000000 ( Similarity of the 1st result )
Disk08/Photos/Mosaic/639069.JPG ( Location of the 2nd result )
1.000000 ( Similarity of the 2nd result )
¡­ ¡­