RUN.md

🏃‍♀️ Inference Code readme

do novel view synthesis given 2D images (on some demo images):

We render azim novel view video by default.

bash scripts/test/demo_view_synthesis.sh

Conduct semantics editing (on some demo images):

bash scripts/test/demo_editing.sh

This script shall output a video with the change of yaw angle and editid scale. To edit a specific attribute on an identity, the following flags need to be set: --smile_ids, --beard_ids, --bangs_ids and --age_ids.

To control the editing scale, set the following flags:

---editing_boundary_scale_upperbound and ---editing_boundary_scale_lowerbound, following the order of Bangs, Smiling, No_Beard and Young. We also have the editing directions of Eyeglass, but this seems to be unstable on StyleSDF and we do not include the results in the final paper.

3D Toonifications with our pre-triaind encoder:

Here we use a fine-tuned 3D GAN with pre-trained E0 encoder (w/o local feature module) to do the inference.

bash scripts/test/demo_toonify.sh

Reproduce the results in Table 1 (Quantitative performance on CelebA-HQ.)

bash scripts/test/eval_2dmetrics_ffhq.sh

---

Render video flags

change the following code to determin some characteristics of the video rendered.

--render_video \ # whether to render video,
--azim_video \ # render azim/ellipsoid video
--video_frames 9 \ # how many frames in the vide 
--no_surface_renderings \ # whether to render the mesh

Misc

Note that for the scripts end with _ada, we use the stage 2 (pure 2D alignment) model to do the inference. It usually yields results with better fidelity but less view consistency. Feel free to try and use the model that suits your need. To download this model, first run python download_ada_models.py to download the pre-trained encoder.