High-Fidelity Synthetic Data Generation Framework for Unique Objects Detection

One of the key barriers to neural network adoption is the lack of computational resources and high-quality training data—particularly for unique objects without existing datasets. This research explores methods for generating realistic synthetic images that preserve the visual properties of target o...

Full description

Saved in:
Bibliographic Details
Main Authors: Nataliya Shakhovska, Bohdan Sydor, Solomiia Liaskovska, Olga Duran, Yevgen Martyn, Volodymyr Vira
Format: Article
Language:English
Published: MDPI AG 2025-05-01
Series:Computation
Subjects:
Online Access:https://www.mdpi.com/2079-3197/13/5/120
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849711164431269888
author Nataliya Shakhovska
Bohdan Sydor
Solomiia Liaskovska
Olga Duran
Yevgen Martyn
Volodymyr Vira
author_facet Nataliya Shakhovska
Bohdan Sydor
Solomiia Liaskovska
Olga Duran
Yevgen Martyn
Volodymyr Vira
author_sort Nataliya Shakhovska
collection DOAJ
description One of the key barriers to neural network adoption is the lack of computational resources and high-quality training data—particularly for unique objects without existing datasets. This research explores methods for generating realistic synthetic images that preserve the visual properties of target objects, ensuring their similarity to real-world appearance. We propose a flexible approach for synthetic data generation, focusing on improved accuracy and adaptability. Unlike many existing methods that rely heavily on specific generative models and require retraining with each new version, our method remains compatible with state-of-the-art models without high computational overhead. It is especially suited for user-defined objects, leveraging a 3D representation to preserve fine details and support integration into diverse environments. The approach also addresses resolution limitations by ensuring consistent object placement within high-quality scenes.
format Article
id doaj-art-e29623a383dd4f7a9b2bc63b4c7df08c
institution DOAJ
issn 2079-3197
language English
publishDate 2025-05-01
publisher MDPI AG
record_format Article
series Computation
spelling doaj-art-e29623a383dd4f7a9b2bc63b4c7df08c2025-08-20T03:14:41ZengMDPI AGComputation2079-31972025-05-0113512010.3390/computation13050120High-Fidelity Synthetic Data Generation Framework for Unique Objects DetectionNataliya Shakhovska0Bohdan Sydor1Solomiia Liaskovska2Olga Duran3Yevgen Martyn4Volodymyr Vira5Department of Artificial Intelligence, Lviv Polytechnic National University, Kniazia Romana Street, 5, 79005 Lviv, UkraineDepartment of Artificial Intelligence, Lviv Polytechnic National University, Kniazia Romana Street, 5, 79005 Lviv, UkraineDepartment of Artificial Intelligence, Lviv Polytechnic National University, Kniazia Romana Street, 5, 79005 Lviv, UkraineDepartment of Mechanical Engineering, Faculty of Engineering, Computing and the Environment, Kingston University, Room RV MB 215, Main Building (RV), Roehampton Vale, Kingston, London KT12EE, UKDepartment of Project Management, Information Technologies and Telecommunication, Lviv State University of Life Safety, 79007 Lviv, UkraineDepartment of Strength of Materials and Structural Mechanics, Lviv Polytechnic National University, 6 Starosolskykh Street, 79013 Lviv, UkraineOne of the key barriers to neural network adoption is the lack of computational resources and high-quality training data—particularly for unique objects without existing datasets. This research explores methods for generating realistic synthetic images that preserve the visual properties of target objects, ensuring their similarity to real-world appearance. We propose a flexible approach for synthetic data generation, focusing on improved accuracy and adaptability. Unlike many existing methods that rely heavily on specific generative models and require retraining with each new version, our method remains compatible with state-of-the-art models without high computational overhead. It is especially suited for user-defined objects, leveraging a 3D representation to preserve fine details and support integration into diverse environments. The approach also addresses resolution limitations by ensuring consistent object placement within high-quality scenes.https://www.mdpi.com/2079-3197/13/5/120neural networksdiffusion generative modelssynthetic training data3D reconstruction
spellingShingle Nataliya Shakhovska
Bohdan Sydor
Solomiia Liaskovska
Olga Duran
Yevgen Martyn
Volodymyr Vira
High-Fidelity Synthetic Data Generation Framework for Unique Objects Detection
Computation
neural networks
diffusion generative models
synthetic training data
3D reconstruction
title High-Fidelity Synthetic Data Generation Framework for Unique Objects Detection
title_full High-Fidelity Synthetic Data Generation Framework for Unique Objects Detection
title_fullStr High-Fidelity Synthetic Data Generation Framework for Unique Objects Detection
title_full_unstemmed High-Fidelity Synthetic Data Generation Framework for Unique Objects Detection
title_short High-Fidelity Synthetic Data Generation Framework for Unique Objects Detection
title_sort high fidelity synthetic data generation framework for unique objects detection
topic neural networks
diffusion generative models
synthetic training data
3D reconstruction
url https://www.mdpi.com/2079-3197/13/5/120
work_keys_str_mv AT nataliyashakhovska highfidelitysyntheticdatagenerationframeworkforuniqueobjectsdetection
AT bohdansydor highfidelitysyntheticdatagenerationframeworkforuniqueobjectsdetection
AT solomiialiaskovska highfidelitysyntheticdatagenerationframeworkforuniqueobjectsdetection
AT olgaduran highfidelitysyntheticdatagenerationframeworkforuniqueobjectsdetection
AT yevgenmartyn highfidelitysyntheticdatagenerationframeworkforuniqueobjectsdetection
AT volodymyrvira highfidelitysyntheticdatagenerationframeworkforuniqueobjectsdetection