output resolution

[mwelponer]

Hello,

I am using parts of your code written in depth_estimation.
nyu_data training is made of pics 640x480. In your predict code, after loading the rgb images you resize them to 640x480 (test.py line 38).

Question 1: what happens if I skip the resize and do the predict with input images with higher resolution (higher than the resolution of the images used for training)?
I mean, I am wondering if trying to infer depths using a network trained using lower resolution images may give good results

Another thing I have noticed is that the output has resolution 320x240. I have found the place where you set the shape of the depth (data.py line 24 for train and line 74 for test).

Question 2: Why half the resolution?
Question 3: What if I want to get depth images with the same resolution of the input?
I have tried to play a bit with the code but I couldn't find out what and where I need to change. Could you give me some hints?

thank you for your help