Exploring the Potential of Synthetic Data to Replace Real Data
Not on the map. Why?
Abstract (arXiv)
The potential of synthetic data to replace real data creates a huge demand for synthetic data in data-hungry AI. This potential is even greater when synthetic data is used for training along with a small number of real images from domains other than the test domain. We find that this potential varies depending on (i) the number of cross-domain real images and (ii) the test set on which the trained model is evaluated. We introduce two new metrics, the train2test distance and APt2t, to evaluate the ability of a cross-domain training set using synthetic data to represent the characteristics of test instances in relation to training performance. Using these metrics, we delve deeper into the factors that influence the potential of synthetic data and uncover some interesting dynamics about how synthetic data impacts training performance. We hope these discoveries will encourage more widespread use of synthetic data.
Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 30 of its 44 references that are in the map. First posted on arXiv 26 Aug 2024.
Builds on
30 of 44 references.
References · 30
- 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
- 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
- 2020Language Models are Few-Shot Learnerscited by 32,419
- 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
- 2014Microsoft COCO: Common Objects in Contextcited by 17,608
- 2017Mask R-CNNcited by 7,277
- 2014Fully Convolutional Networks for Semantic Segmentationcited by 4,859
- 2016Improved Techniques for Training GANscited by 2,970
- 2019Expressive Body Capture: 3D Hands, Face, and Body from a Single Imagecited by 1,612
- 2019AMASS: Archive of Motion Capture as Surface Shapescited by 1,236
- 2020Deep Learning for Person Re-identification: A Survey and Outlookcited by 549
- 2020Detection and Tracking Meet Drones Challengecited by 263
18 more
- 2023BEDLAM: A Synthetic Dataset of Bodies Exhibiting Detailed Lifelike Animated Motioncited by 236
- 2021AGORA: Avatars in Geography Optimized for Regression Analysiscited by 205
- 2023Scaling Laws of Synthetic Images for Model Training ... for Nowcited by 108
- 2022ObjectFolder 2.0: A Multisensory Object Dataset for Sim2Real Transfercited by 75
- 2017Okutama-Action: An Aerial View Video Dataset for Concurrent Human Action Detectioncited by 46
- 2024Score-Guided Diffusion for 3D Human Recoverycited by 39
- 2023LoGoPrompt: Synthetic Text Images Can Be Good Visual Prompts for Vision-Language Modelscited by 29
- 2024Robust Synthetic-to-Real Transfer for Stereo Matchingcited by 15
- 2024RadSimReal: Bridging the Gap Between Synthetic and Real Data in Radar Object Detection With Simulationcited by 14
- 2022Progressive Transformation Learning for Leveraging Virtual Images in Trainingcited by 10
- 2023UAV-Sim: NeRF-based Synthetic Data Generation for UAV-based Perceptioncited by 10
- 2022Archangel: A Hybrid UAV-based Human Detection Benchmark with Position and Pose Metadatacited by 9
- 2024Bridging the Synthetic-to-Authentic Gap: Distortion-Guided Unsupervised Domain Adaptation for Blind Image Quality Assessmentcited by 9
- 2024Diversifying Human Pose in Synthetic Data for Aerial-view Human Detectioncited by 5
- 2023WinSyn: A High Resolution Testbed for Synthetic Datacited by 3
- 2024TK-Planes: Tiered K-Planes with High Dimensional Feature Vectors for Dynamic UAV-based Scenescited by 3
- 2024Exploring the Impact of Synthetic Data for Aerial-view Human Detectioncited by 3
- 2024HUNTER: Unsupervised Human-centric 3D Detection via Transferring Knowledge from Synthetic Instances to Real Scenescited by 2
Built on by
Citing papers per year: 2025: 1; 2026: 2.
3 papers.