GIRAFFE HD: A High-Resolution 3D-aware Generative Model

Project Page <br> Paper

Usage

Create and activate conda environment 'giraffehd': <br>

conda env create -f environment.yml
conda activate giraffehd

Train

Create lmdb dataset: <br>

python prepare_data.py --out LMDB_PATH --n_worker N_WORKER --size SIZE DATASET_PATH

This will convert images to jpeg and pre-resizes them. <br>

Train model in distributed settings:

python -m torch.distributed.launch --nproc_per_node=N_GPU --master_port=PORT train.py \
--wandb --batch BATCH_SIZE --dataset DATASET --size SIZE --datasize DATASIZE LMDB_PATH

Evaluate

Evaluate trained model: <br>

python eval.py --ckpt CKPT --batch BATCH_SIZE --control_i CONTROL_I

Use --control_i to specify which feature to control, <br>

0: fg_shape; 1: fg_app; 2: bg_shape; 3: bg_app; 4: camera rotation angle; 5: elevation angle;
7: scale; 8: translation; 9: rotation;

Change L168-183 in eval.py to specify interpolation interval if needed (training intervals will be used if not specified). For example, set --control_i to 8, and

args.translation_range_min = [0., 0., -0.1]
args.translation_range_max = [0., 0., 0.1]

to perform object vertical translation.

Checkpoints

Model checkpoints are available in google drive.

Acknowledgment

Thanks to giraffe and stylegan2-pytorch

License

This repository is released under the MIT license.

Citation

@inproceedings{xue2022giraffehd,
author    = {Yang Xue and Yuheng Li and Krishna Kumar Singh and Yong Jae Lee},
title     = {GIRAFFE HD: A High-Resolution 3D-aware Generative Model},
booktitle   = {CVPR},
year      = {2022},
}