search

sacmehta / ESPNet

ESPNet: Efficient Spatial Pyramid of Dilated Convolutions for Semantic Segmentation
https://sacmehta.github.io/ESPNet/
MIT License
538 stars 111 forks source link

Is this on purpose?myTransforms.Scale(1280, 720) not (1280,640) #58

Closed wangyihuan123 closed 5 years ago

wangyihuan123 commented 5 years ago

Hi, I notice that the code below, and am wondering whether the scale 720 is on purpose (not 512*1.25)

train/main.py: https://github.com/sacmehta/ESPNet/blob/afe71c38edaee3514ca44e0adcafdf36109bf437/train/main.py#L227

trainDataset_scale2 = myTransforms.Compose([ myTransforms.Normalize(mean=data['mean'], std=data['std']), myTransforms.Scale(1280, 720), # <<<<<<<<<< 512*1.25 == 640 ? myTransforms.RandomCropResize(100), myTransforms.RandomFlip(),

myTransforms.RandomCrop(64),

    myTransforms.ToTensor(args.scaleIn),
    #

])

sacmehta commented 5 years ago

There is no specific reason behind this choice. We just selected it to have 720p resolution.