Using trained model for custom sized images

alexlee-gk / video_prediction

Stochastic Adversarial Video Prediction

https://alexlee-gk.github.io/video_prediction/

MIT License

303 stars 65 forks source link

Using trained model for custom sized images #38

Open malalejandra opened 5 years ago

malalejandra commented 5 years ago

Hi @alexlee-gk. Thank you for your wonderful code, was quite an interesting journey to dig into it!

I trained models for my own dataset of 64x64 images and wondering if it's possible to use the model to sample 512x512 images? If yes, what changes I should make so that the model become scalable?

Thank you!

blackPython commented 5 years ago

To run for 512x512 image you have to change the enocder and decoder.. Not the architecture but the number layers or step size to account for more number of pixels.