g3dr's Introduction

G3DR: Generative 3D Reconstruction in ImageNet (CVPR 2024)

Code for reproducing f_trigen network training and inference.

Prepare config file

Config file for training on imagenet data is available in ./config/ Please update 'dataset_folder', 'renderer_path', 'save_dir' fields before training.

dataset_folder - path to the imagenet dataset along with depth, you can generate the dataset yourself by downloading imagenet and using an off-the-shelf monocular depth estimation network or download it from here: https://drive.google.com/drive/folders/1yAMr1Us9gD6F5P0lCd5qiouyZ9gT5P_n

renderer_path - path to eg3d library, clone/download https://github.com/NVlabs/eg3d/ and provide the path to the repo here.

save_dir - path to dir to save the training logs.

Requirements

Install Clip repo from https://github.com/openai/CLIP

All the other required libraries should be available to download via pip.

Training

after configuring accelerate use the below command for training

accelerate launch train.py --config config/imagenet_train.yaml

for training with a single GPU without accelerate use

python train.py --config config/imagenet_train.yaml

In case of errors replace "model.module" with "model"

Pre-trained model

Pretrained f_trigen on imagenet can be found here: https://drive.google.com/file/d/1Bg5k3IYquph-cZbWJVW0A4kyyd-t7n-d/view

Pretrained super-resolution model weights can be found here: https://drive.google.com/file/d/1Wsa0bbw_oP80O5DdCyE5qIQGunXxwgt7/view

Visualization

CUDA_VISIBLE_DEVICES=0 python visualize.py --load_model ./checkpoint_generic.pt --config ./config/test.yaml --folder ./images/1/

The expected outputs are present in the output folder.

Bibtex

Please consider citing our paper.

@misc{reddy2024g3dr,
      title={G3DR: Generative 3D Reconstruction in ImageNet}, 
      author={Pradyumna Reddy and Ismail Elezi and Jiankang Deng},
      year={2024},
      eprint={2403.00939},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

g3dr's People