SynthRender and IRIS: Open-Source Framework and Dataset for Bidirectional Sim-Real Transfer in Industrial Object Perception

2026-02-24

AI Summary

This study addresses the high cost of acquiring and annotating real-world data for specialized industrial components, which hinders the robust deployment of deep learning models in semi-controlled environments. To overcome this challenge, we propose SynthRender—an open-source synthetic image generation framework that systematically integrates a Reality-to-Simulation pipeline, which reconstructs 3D assets from real 2D images, with guided domain randomization to enable efficient and low-cost bidirectional sim-to-real transfer. Using this framework, we construct IRIS, a dataset comprising paired real-synthetic samples of 32 industrial part categories exhibiting high inter-class similarity and intra-class diversity. Experiments demonstrate that our approach achieves mAP@50 scores of 99.1%, 98.3%, and 95.3% on public robotics, automotive benchmarks, and IRIS, respectively, significantly outperforming existing methods.

Technology Category

• Computer Vision: Computational Photography, Image & Video Synthesis
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Web Mining and Content Analysis: Web data generation and simulation
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Object perception is fundamental for tasks such as robotic material handling and quality inspection. However, modern supervised deep-learning perception models require large datasets for robust automation under semi-uncontrolled conditions. The cost of acquiring and annotating such data for proprietary parts is a major barrier for widespread deployment. In this context, we release SynthRender, an open source framework for synthetic image generation with Guided Domain Randomization capabilities. Furthermore, we benchmark recent Reality-to-Simulation techniques for 3D asset creation from 2D images of real parts. Combined with Domain Randomization, these synthetic assets provide low-overhead, transferable data even for parts lacking 3D files. We also introduce IRIS, the Industrial Real-Sim Imagery Set, containing 32 categories with diverse textures, intra-class variation, strong inter-class similarities and about 20,000 labels. Ablations on multiple benchmarks outline guidelines for efficient data generation with SynthRender. Our method surpasses existing approaches, achieving 99.1% mAP@50 on a public robotics dataset, 98.3% mAP@50 on an automotive benchmark, and 95.3% mAP@50 on IRIS.

Problem

• object perception
• industrial automation
• data annotation
• synthetic data
• sim-to-real transfer

Innovation

• SynthRender
• Guided Domain Randomization
• Reality-to-Simulation
• Industrial Object Perception
• IRIS Dataset