# Multi-Modal Conditioned High-Resolution Transformer for Urban Electromagnetic Field Map Prediction Download PDF

arXiv: https://arxiv.org/abs/2606.27671 · HTML: https://mlsanity.com/paper/2606.27671/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 9 of its 27 references that are in the map.
- First posted on arXiv 26 Jun 2026.

## Builds on (most cited first, 9 of 9)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://mlsanity.com/paper/1505.04597/) (2015)
- [FiLM: Visual Reasoning with a General Conditioning Layer](https://mlsanity.com/paper/1709.07871/) (2017)
- [Deep High-Resolution Representation Learning for Visual Recognition](https://mlsanity.com/paper/1908.07919/) (2019)
- [RadioUNet: Fast Radio Map Estimation with Convolutional Neural Networks](https://mlsanity.com/paper/1911.09002/) (2019)
- [Dataset of Pathloss and ToA Radio Maps With Localization Application](https://mlsanity.com/paper/2212.11777/) (2022)
- [A Scalable and Generalizable Pathloss Map Prediction](https://mlsanity.com/paper/2312.03950/) (2023)
- [Cellular Network Radio Propagation Modeling with Deep Convolutional Neural Networks](https://mlsanity.com/paper/2110.01848/) (2021)
- [RadioMamba: Breaking the Accuracy-Efficiency Trade-off in Radio Map Construction via a Hybrid Mamba-UNet](https://mlsanity.com/paper/2508.09140/) (2025)

## Abstract (arXiv)

Predicting electromagnetic field (EMF) strength in urban environments is essential for cellular network planning but computationally expensive with physics-based simulators. We propose a multi-conditioned dense prediction framework that generates 500 500 EMF maps from building layout images and antenna configurations. Our architecture uses a High-Resolution Transformer (HRFormer) backbone with two complementary conditioning mechanisms: Feature-wise Linear Modulation (FiLM) injects scalar antenna parameters into all backbone stages, while cross-attention fuses 1-D radiation pattern tokens with spatial features at the deepest stage. We further introduce transmitter-relative spatial channels encoding distance, proximity, and bearing from the antenna, enabling coordinate-consistent test-time augmentation (TTA) that reduces test MAE by 6.3%. To address the prediction difficulty imbalance across EMF maps, we design a composite loss combining masked L1, multi-scale structural similarity (MS-SSIM), and a focal L1 term that upweights high-signal pixels, outperforming individual loss components in all metrics. Our best model achieves a test MAE of 0.0461, a 25.2% improvement over a plain UNet baseline and 31.8% over an HRFormer-only baseline.Do-