CartoonGan-tensorflow
CartoonGan-tensorflow copied to clipboard
mnicnc404
Full resolution images
Hi! I've read your paper, and I want to re-train the model with my own dataset of images. How can I do if I want to train the image in full resolution (e.g. 1920x1080)? Do I have to change the architecture and add more layers?
Hi! I've read your paper, and I want to re-train the model with my own dataset of images. How can I do if I want to train the image in full resolution (e.g. 1920x1080)? Do I have to change the architecture and add more layers?
I think you just need more GPU mem. I made it work for 4K images with 2x 11G cards.
Hi! I've read your paper, and I want to re-train the model with my own dataset of images. How can I do if I want to train the image in full resolution (e.g. 1920x1080)? Do I have to change the architecture and add more layers?
I think you just need more GPU mem. I made it work for 4K images with 2x 11G cards.
Do I need to change any params while running for this image?
Also, If I need to run inference on larger image, do I need to train on same size?
Hi! I've read your paper, and I want to re-train the model with my own dataset of images. How can I do if I want to train the image in full resolution (e.g. 1920x1080)? Do I have to change the architecture and add more layers?
I think you just need more GPU mem. I made it work for 4K images with 2x 11G cards.
Do I need to change any params while running for this image?
Also, If I need to run inference on larger image, do I need to train on same size?
In my personal opinion, I don't think so. I think all you need is more GPU memory. To inference on a 4K image, I need to run it on 2X 1080TI (22G in total). Of course, you can run on CPU if you don't have enough GPU memory, but it will be way much slower.