Closed mr17m closed 1 year ago
Hi,
In the official SALICON dataset, fixation maps are provided in .mat format. In this work, we converted these .mat files into .bmp fixation maps for model training. The specific method is to mark the fixation points according to the coordinates given in the .mat files, using pixels with a gray level of 255 on an all-black background. Then we can obtain image-format binary fixation maps.
Hi,
The specific dict_keys you utilized from the .mat files during the conversion process, such as 'header', 'version', 'globals', 'image', 'resolution', or 'gaze'?
Hi,
The specific dict_keys you utilized from the .mat files during the conversion process, such as 'header', 'version', 'globals', 'image', 'resolution', or 'gaze'?
Hi,
I think it should be gaze.
Hello, In your train script you used
MyDataset
function and one of the arguments of this function isfixation_dir=r'datasets\val\val_fixation/
. My question is that from which source you downloaded fixation maps (bmp files that also are mentioned in the csv files)? This is because the format of the fixation maps that are provided in this link are mat files. Should we do preprocessing with one of your scripts?