Vision Transformers for Efficient Indoor Pathloss Radio Map Prediction

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Mkrtchyan, Rafayel, Ghukasyan, Edvard, Petrosyan, Khoren, Khachatrian, Hrant, Raptis, Theofanis P.
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866908353371832320
author Mkrtchyan, Rafayel
Ghukasyan, Edvard
Petrosyan, Khoren
Khachatrian, Hrant
Raptis, Theofanis P.
author_facet Mkrtchyan, Rafayel
Ghukasyan, Edvard
Petrosyan, Khoren
Khachatrian, Hrant
Raptis, Theofanis P.
contents Indoor pathloss prediction is a fundamental task in wireless network planning, yet it remains challenging due to environmental complexity and data scarcity. In this work, we propose a deep learning-based approach utilizing a vision transformer (ViT) architecture with DINO-v2 pretrained weights to model indoor radio propagation. Our method processes a floor map with additional features of the walls to generate indoor pathloss maps. We systematically evaluate the effects of architectural choices, data augmentation strategies, and feature engineering techniques. Our findings indicate that extensive augmentation significantly improves generalization, while feature engineering is crucial in low-data regimes. Through comprehensive experiments, we demonstrate the robustness of our model across different generalization scenarios.
format Preprint
id arxiv_https___arxiv_org_abs_2412_09507
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Vision Transformers for Efficient Indoor Pathloss Radio Map Prediction
Mkrtchyan, Rafayel
Ghukasyan, Edvard
Petrosyan, Khoren
Khachatrian, Hrant
Raptis, Theofanis P.
Computer Vision and Pattern Recognition
Artificial Intelligence
Networking and Internet Architecture
Indoor pathloss prediction is a fundamental task in wireless network planning, yet it remains challenging due to environmental complexity and data scarcity. In this work, we propose a deep learning-based approach utilizing a vision transformer (ViT) architecture with DINO-v2 pretrained weights to model indoor radio propagation. Our method processes a floor map with additional features of the walls to generate indoor pathloss maps. We systematically evaluate the effects of architectural choices, data augmentation strategies, and feature engineering techniques. Our findings indicate that extensive augmentation significantly improves generalization, while feature engineering is crucial in low-data regimes. Through comprehensive experiments, we demonstrate the robustness of our model across different generalization scenarios.
title Vision Transformers for Efficient Indoor Pathloss Radio Map Prediction
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Networking and Internet Architecture
url https://arxiv.org/abs/2412.09507