Base BackBone, Paper Link: RadioDiff, Code Link: GitHub
PINN Enhanced with Helmholtz Equation, Paper Link: RadioDiff-$k^2$, Code Link: GitHub
Efficiency Enhanced RadioDiff, Paper Link: RadioDiff-Turbo
Indoor RM Construction with Physical Information, Paper Link: iRadioDiff, Code Link: GitHub
3D RM with DataSet, Paper Link: RadioDiff-3D, Code Link: GitHub
Sparse Measurement for RM ISAC, Paper Link: RadioDiff-Inverse
Sparse Measurement for NLoS Localization, Paper Link: RadioDiff-Loc
For more RM information, please visit the repo of Awesome-Radio-Map-Categorized
An intelligent radio-map reconstruction system based on diffusion models. ๐ถโจ
RadioDiff-k2 is an advanced radio-map reconstruction project that leverages conditional diffusion models to generate high-quality radio-coverage maps from sparse measurements. The project serves 5G and 6G network planning, propagation prediction, and network optimization. ๐๐ก
- DPM โ deterministic propagation modeling with high speed and accuracy
- IRT4 โ iterative ray tracing with high-precision prediction
- DPMCARK โ vehicle-aware enhancement for urban mobility scenes ๐๐ก
- Conditional diffusion model built on Swin Transformer
- VAE encoder for compact and efficient representation
- Multi-scale processing for flexible resolution support ๐ง ๐งฉ
- Building layouts for realistic urban environments
- Transmitter positions to capture source attributes
- Vehicle data for dynamic occlusions
- k2 features to encode physical propagation traits ๐๏ธ๐๐๐
Python >= 3.8 CUDA >= 11.0 PyTorch >= 1.12
RadioDiff-k2/
โโโ ๐ configs/ # Configuration files
โ โโโ BSDS_sample_*.yaml # Inference configs
โ โโโ BSDS_train_*.yaml # Training configs
โโโ ๐ง denoising_diffusion_pytorch/ # Diffusion core
โโโ ๐ง lib/ # Utilities
โ โโโ loaders.py # Data loaders
โ โโโ modules.py # Network modules
โโโ ๐พ model/ # Pretrained models
โโโ ๐ inference/ # Inference results
โ โโโ DPMCARK/ # DPMCARK outputs
โ โโโ DPMK/ # DPMK outputs
โ โโโ IRT4K/ # IRT4K outputs
โโโ ๐ metrics/ # Evaluation metrics
โโโ ๐ train_cond_ldm.py # Training script
โโโ ๐ฎ sample_cond_ldm.py # Inference script
โโโ ๐๏ธ train_vae.py # VAE training
โโโ ๐งฎ caculate_k.py # k2 feature computation
โโโ ๐ฏ demo.py # Usage examples
โโโ ๐ฆ requirements.txt # Dependencies
โโโ ๐ README.md # Project docs
RadioMapSeer/
โโโ ๐ png/
โ โโโ buildings_complete/ # Building images 256x256
โ โโโ antennas/ # Transmitter positions 256x256
โ โโโ cars/ # Vehicle information optional
โโโ ๐ gain/
โ โโโ DPM/ # DPM simulation results
โ โโโ IRT4/ # IRT4 simulation results
โ โโโ IRT4_k2_neg_norm/ # k2 feature maps
โโโ ๐ metadata/ # Meta files
# Run the k2 feature computation script
python caculate_k.py# Train the main model
python train_cond_ldm.py --cfg configs/BSDS_train_DPMK.yaml
python train_cond_ldm.py --cfg configs/BSDS_train_DPMCARK.yaml
python train_cond_ldm.py --cfg configs/BSDS_train_IRT4K.yaml# DPMCARK inference python sample_cond_ldm.py --cfg configs/BSDS_sample_DPMCARK.yaml # DPMK inference python sample_cond_ldm.py --cfg configs/BSDS_sample_DPMK.yaml # IRT4K inference python sample_cond_ldm.py --cfg configs/BSDS_sample_IRT4K.yaml