Artificial Intelligence-driven Image Synthesis for Sensor Data

Darshan HASSAN SHIVAKUMAR

Diplomová práce

Abstract:

This thesis investigates generative models and identifies a suitable model for each sensor modality (RGB camera, IR sensor, radar, and laser scanner) using data collected from drone-mounted sensors. The goal is to generate high-quality synthetic images that are difficult to acquire at large scale due to limited flight time, regulatory restrictions, safety concerns, and environmental constraints. A common methodological workflow is defined for all sensor modalities, and the model selection and the evaluation strategies are provided for each modality. However, experimental implementation is conducted only for images collected by the RGB camera.

Abstract:

This thesis investigates generative models and identifies a suitable model for each sensor modality (RGB camera, IR sensor, radar, and laser scanner) using data collected from drone-mounted sensors. The goal is to generate high-quality synthetic images that are difficult to acquire at large scale due to limited flight time, regulatory restrictions, safety concerns, and environmental constraints. A common methodological workflow is defined for all sensor modalities, and the model selection and the evaluation strategies are provided for each modality. However, experimental implementation is conducted only for images collected by the RGB camera.

Keywords

synthetic image generation, generative models, diffusion models, one-shot image generation, drone-mounted sensors, RGB images, thermal images, radar range-doppler maps, LiDAR point clouds, image quality assessment.

Jazyk práce: angličtina

Datum vytvoření / odevzdání či podání práce: 31. 1. 2026

Obhajoba závěrečné práce

• Vedoucí: Stafan Kunze, M.Sc.

JIHOČESKÁ UNIVERZITA V ČESKÝCH BUDĚJOVICÍCH

Přírodovědecká fakulta

Magisterský studijní program / obor:
Artificial Intelligence and Data Science / Artificial Intelligence and Data Science