Some question about code of data_preprocessing and enviroment

[LossNAN]

Hello!I am a student in UESTC,thank you for your share,and I have some question about your code.

Fristly, i have downloaded the dateset of SALICON CHALLENGE(image(jpg),fixation map(png),and fixtion(mat)),but when i ran your code of data_preprocessing,this function:

for currFile in tqdm(listImgFiles): tt = dataRepresentation.Target(os.path.join(pathToImages, currFile + '.jpg'), os.path.join(pathToMaps, currFile + '.mat'), os.path.join(pathToFixationMaps, currFile + '.mat'), dataRepresentation.LoadState.loaded, dataRepresentation.InputType.image, dataRepresentation.LoadState.loaded, dataRepresentation.InputType.saliencyMapMatlab, dataRepresentation.LoadState.unloaded, dataRepresentation.InputType.empty ) why you use two dataset about .mat?(dataset only have one .mat set)

Secondly,the code can only run under the linux and gpu?how about windows on pychram and cpu?I met so many question on enviroment.

Thirdly,question about code of 'dataRepresentation' data = (scipy.io.loadmat(self.filePath)['I'] * 255).astype(np.uint8) we kown that the .mat files have three structs(images,gaze,re),and there is not key ,called 'I',when i ran: self.data = (scipy.io.loadmat(self.filePath)['I'] * 255).astype(np.uint8) KeyError: 'I'

I hope you can reply as soon as possibles. I would be very grateful if you can give a detailed explanation about how to run your code on windows Best wishes! ZHIKUN LIN

[junting]

Hi @13551132330 ,

First, you can ignore the fixation map part, since they were not used for training.
Secondly, we have only tested on Linux environment with GPU. Thirdly, sorry that I never met that error before.

Best, Junting