The repository contains files required to reproduce the results. The three compressed filed are (i) torch_code, (ii) datasets, and (iii) experiments. Detailed files description torch_code The main Pytorch source code used for training/testing is provided in torch_code.tar.gz file. datasets The training/validation/testing datasets have been provided in lmdb format which is ready to use in the code. The datasets in datasets.tar.gz contain:

Main train/validation/test dataset:

Training dataset: data_train_OF-decaying2_f0_1_11_12_2_21_22_3_31_32_FHIT_particle_128_Re52-2D_320000_lmdb.lmdb

Validation dataset: data_valid_outOfSample_OF-decaying2_f0_1_11_12_2_21_22_3_31_32_FHIT_particle_128_Re52-2D_8000_lmdb.lmdb

Test dataset: data_test_outOfSample_OF-decaying2_f0_1_11_12_2_21_22_3_31_32_FHIT_particle_128_Re52-2D_16000_lmdb.lmdb

Note that the samples from 20 DNS cases are collected in order (each case 16000 samples for training and 800 samples for testing) which can be recognized using the provided metadata file in each folder.

Particle-free training and test datasets (used in Fig 6 of the paper):

Particle-free training dataset: data_train_OF-f0_FHIT_particle_128_Re52_prolonged-2D_102400_lmdb.lmdb

Particle-free test dataset: data_test_outOfSample_OF-f0_FHIT_particle_128_Re52_prolonged-2D_800_lmdb.lmdb

Out of sample test datasets:

Test Case4 in the paper: data_test_outOfSample_OF-f41_FHIT_particle_128_Re52_test-2D_800_lmdb.lmdb

Test Case5 in the paper: data_test_outOfSample_OF-f51_FHIT_particle_128_Re52_test-2D_800_lmdb.lmdb

experiments The trained models are provided in experiments.tar.gz file. Each experiment contains the log file of the training, the last training state (for restart) and the model wights used in the publication.

Conditional model:

conditionalSRGAN trained model using particle-free dataset (used in Figs 6 and 7 of the paper): 00110-01G_PFT-NoPrt_ArchT_condSRGANModel_L64SP4x_Gcond_WaveDisc_f256g128b16_I64_BS16x2_Pix1-Grada-Adva_LrG45D5_fixedLR_DS-f0-102k_cPad_20241218

conditionalSRGAN trained model using the main dataset (used in Figs 9-13 and Figs 15-16 of the paper): 01004-00H_PFT-Prt_ArchTest_condSRGANModel_L64SP4x_Gcond_WaveDisc_f256g128b16_I64_BS32x4_Pix1-Grada-Adva_LrG45D5_fixedLR_DS-fxD-320k_cPad_20241219

Traditional model:

unconditional SRGAN model trained model using the main dataset (used in Fig 14 of the paper): 01005-00H_PFT-Prt_DiscTest_condSRGANModel_L64SP4x_Gcond_TradDisc_f256g128b16_I64_BS32x4_Pix1-Grada-Adva_LrG45D5_fixedLR_DS-fxD-320k_cPad_20241224

How to Build the environment To build the environment required for the training and inference you need Anaconda. Go to the torch_code folder and conda env create -f environment.yml

Then create ipython kernel for post processing, conda activate torch_22_2025_Shamooni_POF

python -m ipykernel install --user --name ipyk_torch_22_2025_Shamooni_POF --display-name "ipython kernel for post processing of POF2025"

Perform training It is suggested to create softlinks to the dataset directly in the torch_code folder: cd torch_code ln -s <path to the dataset folder> datasets

Then activate the conda environment conda activate torch_22_2025_Shamooni_POF

An example script to run on single node with 2 GPUs: torchrun --standalone --nnodes=1 --nproc_per_node=2 train.py -opt options/train/condSRGAN/00110-01G_PFT-NoPrt_ArchT.yml --launcher pytorch

Make sure that the paths to datasets "dataroot_gt" and "meta_info_file" for both training and validation data in option files are set correctly.