# Applying Domain Randomization to Synthetic Data for Object Category Detection

arXiv: https://arxiv.org/abs/1807.09834 · HTML: https://mlsanity.com/paper/1807.09834/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 16 Jul 2018.

## Built on by (most cited first, 9 of 9)

- [Invariant Information Bottleneck for Domain Generalization](https://mlsanity.com/paper/2106.06333/) (2021)
- [AKB-48: A Real-World Articulated Object Knowledge Base](https://mlsanity.com/paper/2202.08432/) (2022)
- [Towards Real-World Category-level Articulation Pose Estimation](https://mlsanity.com/paper/2105.03260/) (2021)
- [A Scaling Law for Synthetic-to-Real Transfer: How Much Is Your Pre-training Effective?](https://mlsanity.com/paper/2108.11018/) (2021)
- [Improving Object Detector Training on Synthetic Data by Starting With a Strong Baseline Methodology](https://mlsanity.com/paper/2405.19822/) (2024)
- [Synthetic Data Generation for Bridging Sim2Real Gap in a Production Environment](https://mlsanity.com/paper/2311.11039/) (2023)
- [Synthetic Image Data for Deep Learning](https://mlsanity.com/paper/2212.06232/) (2022)
- [Drone Detection using Deep Neural Networks Trained on Pure Synthetic Data](https://mlsanity.com/paper/2411.09077/) (2024)
- [Synthetic Data and Hierarchical Object Detection in Overhead Imagery](https://mlsanity.com/paper/2102.00103/) (2021)

## Abstract (arXiv)

Recent advances in deep learning-based object detection techniques have revolutionized their applicability in several fields. However, since these methods rely on unwieldy and large amounts of data, a common practice is to download models pre-trained on standard datasets and fine-tune them for specific application domains with a small set of domain relevant images. In this work, we show that using synthetic datasets that are not necessarily photo-realistic can be a better alternative to simply fine-tune pre-trained networks. Specifically, our results show an impressive 25% improvement in the mAP metric over a fine-tuning baseline when only about 200 labelled images are available to train. Finally, an ablation study of our results is presented to delineate the individual contribution of different components in the randomization pipeline.