Synthetic Industrial Object Detection: GenAI vs. Feature-Based Methods

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Araya-Martinez, Jose Moises, Reig, Adrián Sanchis, Mohan, Gautham, Sardari, Sarvenaz, Lambrecht, Jens, Krüger, Jörg
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866917112020205568
author Araya-Martinez, Jose Moises
Reig, Adrián Sanchis
Mohan, Gautham
Sardari, Sarvenaz
Lambrecht, Jens
Krüger, Jörg
author_facet Araya-Martinez, Jose Moises
Reig, Adrián Sanchis
Mohan, Gautham
Sardari, Sarvenaz
Lambrecht, Jens
Krüger, Jörg
contents Reducing the burden of data generation and annotation remains a major challenge for the cost-effective deployment of machine learning in industrial and robotics settings. While synthetic rendering is a promising solution, bridging the sim-to-real gap often requires expert intervention. In this work, we benchmark a range of domain randomization (DR) and domain adaptation (DA) techniques, including feature-based methods, generative AI (GenAI), and classical rendering approaches, for creating contextualized synthetic data without manual annotation. Our evaluation focuses on the effectiveness and efficiency of low-level and high-level feature alignment, as well as a controlled diffusion-based DA method guided by prompts generated from real-world contexts. We validate our methods on two datasets: a proprietary industrial dataset (automotive and logistics) and a public robotics dataset. Results show that if render-based data with enough variability is available as seed, simpler feature-based methods, such as brightness-based and perceptual hashing filtering, outperform more complex GenAI-based approaches in both accuracy and resource efficiency. Perceptual hashing consistently achieves the highest performance, with mAP50 scores of 98% and 67% on the industrial and robotics datasets, respectively. Additionally, GenAI methods present significant time overhead for data generation at no apparent improvement of sim-to-real mAP values compared to simpler methods. Our findings offer actionable insights for efficiently bridging the sim-to-real gap, enabling high real-world performance from models trained exclusively on synthetic data.
format Preprint
id arxiv_https___arxiv_org_abs_2511_23241
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Synthetic Industrial Object Detection: GenAI vs. Feature-Based Methods
Araya-Martinez, Jose Moises
Reig, Adrián Sanchis
Mohan, Gautham
Sardari, Sarvenaz
Lambrecht, Jens
Krüger, Jörg
Computer Vision and Pattern Recognition
68T45
I.2.10
Reducing the burden of data generation and annotation remains a major challenge for the cost-effective deployment of machine learning in industrial and robotics settings. While synthetic rendering is a promising solution, bridging the sim-to-real gap often requires expert intervention. In this work, we benchmark a range of domain randomization (DR) and domain adaptation (DA) techniques, including feature-based methods, generative AI (GenAI), and classical rendering approaches, for creating contextualized synthetic data without manual annotation. Our evaluation focuses on the effectiveness and efficiency of low-level and high-level feature alignment, as well as a controlled diffusion-based DA method guided by prompts generated from real-world contexts. We validate our methods on two datasets: a proprietary industrial dataset (automotive and logistics) and a public robotics dataset. Results show that if render-based data with enough variability is available as seed, simpler feature-based methods, such as brightness-based and perceptual hashing filtering, outperform more complex GenAI-based approaches in both accuracy and resource efficiency. Perceptual hashing consistently achieves the highest performance, with mAP50 scores of 98% and 67% on the industrial and robotics datasets, respectively. Additionally, GenAI methods present significant time overhead for data generation at no apparent improvement of sim-to-real mAP values compared to simpler methods. Our findings offer actionable insights for efficiently bridging the sim-to-real gap, enabling high real-world performance from models trained exclusively on synthetic data.
title Synthetic Industrial Object Detection: GenAI vs. Feature-Based Methods
topic Computer Vision and Pattern Recognition
68T45
I.2.10
url https://arxiv.org/abs/2511.23241