search

brownvc / fast-synth

Public code release for our CVPR 2019 paper "Fast and Flexible Indoor Scene Synthesis via Deep Convolutional Generative Models"
Other
54 stars 20 forks source link

Do the input-pics have different dimensions sometimes? #3

Open xpp2000 opened 3 years ago

xpp2000 commented 3 years ago

I am confused. each input-pic has C+8 channels, but different input-pic may has different C(categories). Are the numbers of input-pics various? If yes, i have no idea how to train the model, such as cat.model which declare a fixed input channel. Actually，i infer that the C is whole number of categories even though a input-pic does not have some objects. If you can help me to solve this problem, I would be very grateful. thank you.

kwang-ether commented 3 years ago

C depends on room type, which is why we load the frequency files for the specific room types the number of channels are the same for all room images of the same room type