/MorphFlow

[ECCV 2022] Multiview Regenerative Morphing with Dual Flows

Primary LanguagePython

MorphFlow

Multiview Regenerative Morphing with Dual Flows (Project Page, Paper).

Installation

git clone https://github.com/jimtsai23/MorphFlow.git
cd MorphFlow
pip install -r requirements.txt

Pytorch and torch_scatter installation is machine dependent. Please install the correct version for your machine.

Implementation

Users can use collected datasets or their own images as input. For instance, to morph between Lego and Chair provided in Synthetic-NeRF dataset, users can optimize the explicit representation of each scene with

$ python run.py --config configs/nerf/lego.py
$ python run.py --config configs/nerf/chair.py

Then, to morph from Lego to Chair, the optimization is done with

$ python run.py --config configs/nerf/lego.py --morph lego,chair,weight

where weight specifies blending weight of the source, or Lego in this case.