View in EDS

ENERGY EFFICIENCY IN PHOTOGRAMMETRY: A COMPARATIVE ANALYSIS OF DATASET, HARDWARE, AND RESOLUTION EFFECTS IN AGISOFT METASHAPE

Saved in:
Bibliographic Details
Title: ENERGY EFFICIENCY IN PHOTOGRAMMETRY: A COMPARATIVE ANALYSIS OF DATASET, HARDWARE, AND RESOLUTION EFFECTS IN AGISOFT METASHAPE
Authors: Kolarek, Branimir, Davidović, Davor, Maričević, Marko
Source: Proceedings 28th International conference on printing, design and graphic communication. :28-40
Publisher Information: 2025.
Publication Year: 2025
Subject Terms: Structure from Motion (SfM), processing time, Agisoft Metashape, dataset quality, energy consumption, photogrammetry, hardware comparison, energy efficiency, image resolution
Description: Photogrammetry enables detailed 3D reconstruction from images but often involves computationally intensive workflows. While processing time is frequently analysed, energy consumption is a critical but less quantified factor, particularly concerning variations in datasets, hardware, and processing parameters. This study quantitatively evaluates the impact of dataset type (uniform Single-Source vs. heterogeneous Multi-Source), hardware platform (energy-efficient Apple Mac Mini M4 vs. high-performance Windows/NVIDIA RTX 4090), and input image resolution (4000px, 6000px, Native) on processing time and total energy consumption using Agisoft Metashape Professional. Experiments involved processing two 335-image datasets across the hardware and resolution configurations using a standard workflow with consistent ‘Medium’ quality settings, while monitoring time and energy usage. Results indicate the high-performance system was significantly faster (average 1.74x) but consumed substantially more energy (average 4.65x) than the energy-efficient platform. Energy consumption scaled approximately linearly with the total number of aligned pixels processed under constant settings. The Single-Source dataset demonstrated greater robustness, succeeding where the Multi-Source dataset failed in one low-resource scenario, although the Multi-Source dataset showed competitive or better efficiency in some successful tests on the energy-efficient hardware. These findings highlight crucial trade-offs between speed, energy efficiency, and robustness, providing empirical data to inform workflow optimisation based on specific project constraints.
Document Type: Conference object
ISSN: 3102-2324
Accession Number: edsair.dris...01492..dd4cc2f508fe9493047e51a13d813f48
Database: OpenAIRE
Description
Abstract:Photogrammetry enables detailed 3D reconstruction from images but often involves computationally intensive workflows. While processing time is frequently analysed, energy consumption is a critical but less quantified factor, particularly concerning variations in datasets, hardware, and processing parameters. This study quantitatively evaluates the impact of dataset type (uniform Single-Source vs. heterogeneous Multi-Source), hardware platform (energy-efficient Apple Mac Mini M4 vs. high-performance Windows/NVIDIA RTX 4090), and input image resolution (4000px, 6000px, Native) on processing time and total energy consumption using Agisoft Metashape Professional. Experiments involved processing two 335-image datasets across the hardware and resolution configurations using a standard workflow with consistent ‘Medium’ quality settings, while monitoring time and energy usage. Results indicate the high-performance system was significantly faster (average 1.74x) but consumed substantially more energy (average 4.65x) than the energy-efficient platform. Energy consumption scaled approximately linearly with the total number of aligned pixels processed under constant settings. The Single-Source dataset demonstrated greater robustness, succeeding where the Multi-Source dataset failed in one low-resource scenario, although the Multi-Source dataset showed competitive or better efficiency in some successful tests on the energy-efficient hardware. These findings highlight crucial trade-offs between speed, energy efficiency, and robustness, providing empirical data to inform workflow optimisation based on specific project constraints.
ISSN:31022324