fmu2 / 3d_photo_stylization Goto Github PK

How to solve this problem?

I ran

python test_ldi.py --model_path ckpt/r3/stylize.pth --ldi_path samples/ldi/unsplash/00042.mat --style_path samples/style/7.jpg

is there anything wrong?

I ran the unsplash/00042.may with style 14.jpg. I changed the code to see what the code to see the values in the .mat file. Here they are:

LDI loaded: fov=60.0 h=640 w=960 rgb.shape=(1094154, 3) uv.shape=(1094154, 2) z.shape=(1094154, 1)

These do not match the details given in issue #3 . While the datasets are now available, the code needed to pre-process them and the code needed to use test_ldi on own own images (based most likely on 3dinpaint) are need such that we can do validation. This work is really amazing, but simple things that are missing make it very hard to even get started with own datasets or photos.

Hi, researchers,

Thanks for sharing this fantastic project. I want to train the model and find the link to the pre-processed coco2014 dataset (https://drive.google.com/drive/folders/1qtv-fBrCJXT93w0QoFbt8YqeI09pS8EM?usp=sharing) is broken (404 error) and also missing the files test_ldi_render.py and test_ldi_model.py.

Appreciate for repairing the broken link, and thanks in advance.

First, the ratio of aspect is not correct.
Second, a black area exists in the boundary of image.
Third, the quality of stylization is not perfect as in your demo.

I want to know how to solve it, thank you.

The detail of LeReS is better than Mono. But because of the plentiful detail, pixels are more cluttered on the render result. Such as the resdual trace of hairs or the broken limbs. I think it's not normal. How can I use the depths of LeReS to be better?

Hello
I saw the LPIPS metric in the metric.py file. I have seen three other metrics in the trainer.py file, but have not seen how to use the LPIPS metric. Could you provide that? Thank you.

Hello,

In order to be able to reproduce or continue research based on your work, the pre-processed datasets are needed. Please correct the links at your earliest convenience. Thank you!

I had follow facebook' 3D photo based on glb model,there we can get arbitrary views from the 3D model interacting with people .we can experience facebook' 3D photo in fb‘ app already.
any compared?

Hi! The output of the 3d photo inpainting stage is a .ply file. I was wondering how I can create a .mat file out of it?

Hi! Is it possible to run this on Windows 10?