nahidalam/mpg2

Official code for MPG2: Multi-attribute Pizza Generator: Cross-domain Attribute Control with Conditional StyleGAN

This is the official code for Multi-attribute Pizza Generator (MPG2): Cross-domain Attribute Control with Conditional StyleGAN.

Paper	Demo

Setup Environment

can NOT run on CPU

conda create -n mpg python=3.8
conda activate mpg
git clone [email protected]:klory/food_project.git
cd food_project
pip install -r requirements.txt
pip install git+https://github.com/pytorch/tnt.git@master

Pretrained models

Pretrained models are stored in google-link, files are already in their desired locations, so following the same directory structure will minimize burdens to run the code inside the project (some files are not necessary for the current version of the project as of 2021-03-31).

Pizza10 dataset

Please follow MPG repository.

Ingredient classifier

Please follow MPG repository.

PizzaView dataset

Download PizzaView Dataset from google-link/data/Pizza3D.

cd to datasets/

$ python pizza3d.py

View regressor

cd to view_regressor/

Train

$ CUDA_VISIBLE_DEVICES=0 python train.py --wandb=0

Validate

Download the pretrained model google-link/view_regressor/runs/pizza3d/1ab8hru7/00004999.ckpt:

$ CUDA_VISIBLE_DEVICES=0 python val.py --ckpt_path=/runs/pizza3d/1ab8hru7/00004999.ckpt

MPG2

cd to mpg/,

Train

$ CUDA_VISIBLE_DEVICES=0,1 python train.py --wandb=0

Validate

Download the pretrained model google-linkmpg/runs/30cupu9m/00260000.ckpt.

cd to metrics/:

CUDA_VISIBLE_DEVICES=0 python generate_samples.py --model=mpg

---

Metrics

cd to metrics/,

For more about FID and mAP, follow MPG repository.

FID (Frechet Inception Distance)

To compute FID, we need to first compute the statistics of the real images.

CUDA_VISIBLE_DEVICES=0 python calc_inception.py

then

$ CUDA_VISIBLE_DEVICES=0 python fid.py --model=mpg

I got FID=6.33 using the provided checkpoint under 5000 samples, and FID=4.84 under 50000 samples.

mAE (mean Absolute Error) for view attributes

Computing mAE uses the pre-trained view regressor.

$ CUDA_VISIBLE_DEVICES=0 python mAE.py --model=mpg

Demo

cd to metrics/.

CUDA_VISIBLE_DEVICES=0 streamlit run app.py