aiarc 0.0.2

Chemical and Pharmaceutical Autoencoder - Providing reproducible modelling for quantum chemistry

aiarc

Why

I prototyped novel and good algorithms for chemistry in 2021 and it took me two years to get them publication ready. Of course I did other stuff like thesis, work, exams etc. However, the development time was far too long.

What

Preliminaries and Learnings

• I need my own framework
• I learned from aiarc that I need more flexibility for the major parts
• I need a directory structure and saving structure to properly distribute the files
• there could be an initializer
• methods should be imported easier through the init files
• Testing should be more rigorous (in each new project)

The package is aimed to investigate algorithms faster

The focus is on making algorithms and sparing out the rest

Thus the focus is on an easy interface for investigation. PytorchGeometric gives a good guideline on how to do this. It is more though, and also about producing plots to score the overall algorithm.

aiarc is aimed to improve rapid prototyping of AI algorithms on graphs (timeseries, molecules, networks etc) for building novel simulation applications fast and at scale.

The package shall provide production grade algorithms in an automated way but with user interaction.

Lessons learned from other projects

The lessons I learned from other projects were

• Never trust propietary software that is used for politics, e.g. Microsoft, Nvidia drivers, Canonical etc
• Avoid trusting open-source software that is aimed to upsell a paid software distribution (e.g. Canonical, or security software)
• Code high quality instead of coding fast -> is faster in the end
• Try to code consistently instead of a lot at once

How it should work

Work on a project base -> provide a framework to start a project A project needs differnt things

• Configuration
• training
• Monitoring
• Logging
• Analysis

Foundations

I am having some years of experience now, trying to model in an regulatory environment such as drug design, risk assessement, health care and chemistry.

Packages I have built in that context are

• chembee
• aiarc

An architectural pattern became prevalent such that I found it in torchsr, too (which had mutiple years of development, too)

• torchsr

Now, with aiarc we aim to get modelling to the next stage.

Packages to use

• jax
• pytorch-gym (?)
• torch

torchsr

Super-Resolution Networks for Pytorch

Super-resolution is a process that increases the resolution of an image, adding additional details. Methods using neural networks give the most accurate results, much better than other interpolation methods. With the right training, it is even possible to make photo-realistic images.

For example, here is a low-resolution image, magnified x4 by a neural network, and a high resolution image of the same object:

In this repository, you will find:

• the popular super-resolution networks, pretrained
• common super-resolution datasets
• a unified training script for all models

Models

The following pretrained models are available. Click on the links for the paper:

• EDSR
• CARN
• RDN
• RCAN
• NinaSR

Newer and larger models perform better: the most accurate models are EDSR (huge), RCAN and NinaSR-B2. For practical applications, I recommend a smaller model, such as NinaSR-B1.

Expand benchmark results

Set5 results

Network	Parameters (M)	2x (PSNR/SSIM)	3x (PSNR/SSIM)	4x (PSNR/SSIM)
carn	1.59	37.88 / 0.9600	34.32 / 0.9265	32.14 / 0.8942
carn_m	0.41	37.68 / 0.9594	34.06 / 0.9247	31.88 / 0.8907
edsr_baseline	1.37	37.98 / 0.9604	34.37 / 0.9270	32.09 / 0.8936
edsr	40.7	38.19 / 0.9609	34.68 / 0.9293	32.48 / 0.8985
ninasr_b0	0.10	37.72 / 0.9594	33.96 / 0.9234	31.77 / 0.8877
ninasr_b1	1.02	38.14 / 0.9609	34.48 / 0.9277	32.28 / 0.8955
ninasr_b2	10.0	38.21 / 0.9612	34.61 / 0.9288	32.45 / 0.8973
rcan	15.4	38.27 / 0.9614	34.76 / 0.9299	32.64 / 0.9000
rdn	22.1	38.12 / 0.9609	33.98 / 0.9234	32.35 / 0.8968

Set14 results

Network	Parameters (M)	2x (PSNR/SSIM)	3x (PSNR/SSIM)	4x (PSNR/SSIM)
carn	1.59	33.57 / 0.9173	30.30 / 0.8412	28.61 / 0.7806
carn_m	0.41	33.30 / 0.9151	30.10 / 0.8374	28.42 / 0.7764
edsr_baseline	1.37	33.57 / 0.9174	30.28 / 0.8414	28.58 / 0.7804
edsr	40.7	33.95 / 0.9201	30.53 / 0.8464	28.81 / 0.7872
ninasr_b0	0.10	33.24 / 0.9144	30.02 / 0.8355	28.28 / 0.7727
ninasr_b1	1.02	33.71 / 0.9189	30.41 / 0.8437	28.71 / 0.7840
ninasr_b2	10.0	34.00 / 0.9206	30.53 / 0.8461	28.80 / 0.7863
rcan	15.4	34.13 / 0.9216	30.63 / 0.8475	28.85 / 0.7878
rdn	22.1	33.71 / 0.9182	30.07 / 0.8373	28.72 / 0.7846

DIV2K results (validation set)

Network	Parameters (M)	2x (PSNR/SSIM)	3x (PSNR/SSIM)	4x (PSNR/SSIM)	8x (PSNR/SSIM)
carn	1.59	36.08 / 0.9451	32.37 / 0.8871	30.43 / 0.8366	N/A
carn_m	0.41	35.76 / 0.9429	32.09 / 0.8827	30.18 / 0.8313	N/A
edsr_baseline	1.37	36.13 / 0.9455	32.41 / 0.8878	30.43 / 0.8370	N/A
edsr	40.7	36.56 / 0.9485	32.75 / 0.8933	30.73 / 0.8445	N/A
ninasr_b0	0.10	35.77 / 0.9428	32.06 / 0.8818	30.09 / 0.8293	26.60 / 0.7084
ninasr_b1	1.02	36.35 / 0.9471	32.51 / 0.8892	30.56 / 0.8405	26.96 / 0.7207
ninasr_b2	10.0	36.52 / 0.9482	32.73 / 0.8926	30.73 / 0.8437	27.07 / 0.7246
rcan	15.4	36.61 / 0.9489	32.78 / 0.8935	30.73 / 0.8447	27.17 / 0.7292
rdn	22.1	36.32 / 0.9468	32.04 / 0.8822	30.61 / 0.8414	N/A

B100 results

Network	Parameters (M)	2x (PSNR/SSIM)	3x (PSNR/SSIM)	4x (PSNR/SSIM)
carn	1.59	32.12 / 0.8986	29.07 / 0.8042	27.58 / 0.7355
carn_m	0.41	31.97 / 0.8971	28.94 / 0.8010	27.45 / 0.7312
edsr_baseline	1.37	32.15 / 0.8993	29.08 / 0.8051	27.56 / 0.7354
edsr	40.7	32.35 / 0.9019	29.26 / 0.8096	27.72 / 0.7419
ninasr_b0	0.10	31.97 / 0.8974	28.90 / 0.8000	27.36 / 0.7290
ninasr_b1	1.02	32.24 / 0.9004	29.13 / 0.8061	27.62 / 0.7377
ninasr_b2	10.0	32.32 / 0.9014	29.23 / 0.8087	27.71 / 0.7407
rcan	15.4	32.39 / 0.9024	29.30 / 0.8106	27.74 / 0.7429
rdn	22.1	32.25 / 0.9006	28.90 / 0.8004	27.66 / 0.7388

Urban100 results

Network	Parameters (M)	2x (PSNR/SSIM)	3x (PSNR/SSIM)	4x (PSNR/SSIM)
carn	1.59	31.95 / 0.9263	28.07 / 0.849	26.07 / 0.78349
carn_m	0.41	31.30 / 0.9200	27.57 / 0.839	25.64 / 0.76961
edsr_baseline	1.37	31.98 / 0.9271	28.15 / 0.852	26.03 / 0.78424
edsr	40.7	32.97 / 0.9358	28.81 / 0.865	26.65 / 0.80328
ninasr_b0	0.10	31.33 / 0.9204	27.48 / 0.8374	25.45 / 0.7645
ninasr_b1	1.02	32.48 / 0.9319	28.29 / 0.8555	26.25 / 0.7914
ninasr_b2	10.0	32.91 / 0.9354	28.70 / 0.8640	26.54 / 0.8008
rcan	15.4	33.19 / 0.9372	29.01 / 0.868	26.75 / 0.80624
rdn	22.1	32.41 / 0.9310	27.49 / 0.838	26.36 / 0.79460

All models are defined in torchsr.models. Other useful tools to augment your models, such as self-ensemble methods and tiling, are present in torchsr.models.utils.

Datasets

The following datasets are available. Click on the links for the project page:

• DIV2K
• RealSR
• Flicr2K
• REDS
• Set5, Set14, B100, Urban100

All datasets are defined in torchsr.datasets. They return a list of images, with the high-resolution image followed by downscaled or degraded versions. Data augmentation methods are provided in torchsr.transforms.

Datasets are downloaded automatically when using the download=True flag, or by running the corresponding script i.e. ./scripts/download_div2k.sh.

Usage

from torchsr.datasets import Div2K
from torchsr.models import ninasr_b0
from torchvision.transforms.functional import to_pil_image, to_tensor

# Div2K dataset
dataset = Div2K(root="./data", scale=2, download=False)

# Get the first image in the dataset (High-Res and Low-Res)
hr, lr = dataset[0]

# Download a pretrained NinaSR model
model = ninasr_b0(scale=2, pretrained=True)

# Run the Super-Resolution model
lr_t = to_tensor(lr).unsqueeze(0)
sr_t = model(lr_t)
sr = to_pil_image(sr_t.squeeze(0))
sr.show()

Expand more examples

from torchsr.datasets import Div2K
from torchsr.models import edsr, rcan
from torchsr.models.utils import ChoppedModel, SelfEnsembleModel
from torchsr.transforms import ColorJitter, Compose, RandomCrop

# Div2K dataset, cropped to 256px, width color jitter
dataset = Div2K(
    root="./data", scale=2, download=False,
    transform=Compose([
        RandomCrop(256, scales=[1, 2]),
        ColorJitter(brightness=0.2)
    ]))

# Pretrained RCAN model, with tiling for large images
model = ChoppedModel(
    rcan(scale=2, pretrained=True), scale=2,
    chop_size=400, chop_overlap=10)

# Pretrained EDSR model, with self-ensemble method for higher quality
model = SelfEnsembleModel(edsr(scale=2, pretrained=True))

Training

A script is available to train the models from scratch, evaluate them, and much more. It is not part of the pip package, and requires additional dependencies. More examples are available in scripts/.

pip install piq tqdm tensorboard  # Additional dependencies
python -m torchsr.train -h
python -m torchsr.train --arch edsr_baseline --scale 2 --download-pretrained --images test/butterfly.png --destination results/
python -m torchsr.train --arch edsr_baseline --scale 2 --download-pretrained --validation-only
python -m torchsr.train --arch edsr_baseline --scale 2 --epochs 300 --loss l1 --dataset-train div2k_bicubic

You can evaluate models from the command line as well. For example, for EDSR with the paper's PSNR evaluation:

python -m torchsr.train --validation-only --arch edsr_baseline --scale 2 --dataset-val set5 --chop-size 400 --download-pretrained --shave-border 2 --eval-luminance

Acknowledgements

Thanks to the people behind torchvision and EDSR, whose work inspired this repository. Some of the models available here come from EDSR-PyTorch and CARN-PyTorch.

To cite this work, please use:

@misc{torchsr,
  author = {Gabriel Gouvine},
  title = {Super-Resolution Networks for Pytorch},
  year = {2021},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/Coloquinte/torchSR}},
  doi = {10.5281/zenodo.4868308}
}

@misc{ninasr,
  author = {Gabriel Gouvine},
  title = {NinaSR: Efficient Small and Large ConvNets for Super-Resolution},
  year = {2021},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/Coloquinte/torchSR/blob/main/doc/NinaSR.md}},
  doi = {10.5281/zenodo.4868308}
}