Generative AI holds promise for advancing human factors and ergonomics. However, limited training data can reduce variability in generative models. This study investigates using transfer learning to enhance variability in generative posture prediction. We pre-trained a conditional diffusion model on lifting postures where hands are near body center and fine-tuned it on limited extended-reach postures. Compared to training from scratch, transfer learning significantly improved joint angle variability across multiple body segments while maintaining similar accuracy in posture similarity and validity. Additionally, it reduced training time by over 90%, demonstrating efficiency benefits. These findings highlight transfer learning’s potential to enrich generative model outputs with more variable ergonomics data, supporting scalable and adaptive workplace design.
ChenM.FangY.WangX.LuoH.GengY.ZhangX.HuangC.LiuW.WangB. (2020). Diversity transfer network for few-shot learning. Proceedings of the AAAI Conference on Artificial Intelligence, 34(07), 10559–10566.
2.
ChenX.XuC.YangX.SongL.TaoD. (2018). Gated-GAN: Adversarial gated networks for multi-collection style transfer. IEEE Transactions on Image Processing, 28(2), 546–560.
3.
DrljevicO. (2024). Automated ergonomics correction with neural networks. Technische Universität Wien.
4.
FanC.MeiQ.LiX. (2024). Assisting in the identification of ergonomic risks for workers: A large vision-language model approach. ISARC. Proceedings of the International Symposium on Automation and Robotics in Construction, 41, 1010–1017.
5.
GatysL. A.EckerA. S.BethgeM. (2016). Image style transfer using convolutional neural networks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 2414–2423). IEEE. https://doi.org/10.1109/CVPR.2016.265
6.
HansA.WenY.JainN.KirchenbauerJ.KazemiH.SinghaniaP.SinghS.SomepalliG.GeipingJ.BhateleA. (2024). Be like a goldfish, don’t memorize! Mitigating memorization in generative LLMs. arXiv preprint arXiv:2406.10209.
7.
LiL.PrabhuS.XieZ.WangH.LuL.XuX. (2021). Lifting posture prediction with generative models for improving occupational safety. IEEE Transactions on Human-Machine Systems, 51(5), 494–503.
8.
LyuP.BaiX.YaoC.ZhuZ.HuangT.LiuW. (2017). Auto-encoder guided GAN for Chinese calligraphy synthesis. 2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR), 1, 1095–1100.
9.
MacwanN.HudeA. J.IyerH.JeongH.GuoS. (2024). High-fidelity worker motion simulation with generative AI. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 68(1), 1540–1541.
10.
OjhaU.LiY.LuJ.EfrosA. A.LeeY. J.ShechtmanE.ZhangR. (2021). Few-shot image generation via cross-domain correspondence. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 10743–10752.
11.
ÖzkanoğluM. A.OzerS. (2022). InfraGAN: A GAN architecture to transfer visible images to infrared domain. Pattern Recognition Letters, 155, 69–76.
12.
ParikhP.PenfieldJ.BarkerR.McGowanB.MallonJ. R. (2024). NLP-based ergonomics MSD risk root cause analysis and risk controls recommendation. Ergonomics, 68(7), 1124–1136.
13.
QingL.SuB.JungS.LuL.WangH.XuX. (2024). Predicting human postures for manual material handling tasks using a conditional diffusion model. IEEE Transactions on Human-Machine Systems, 54(6), 723–732.
14.
QingL.SuB.XieZ.JungS.LuL.WangH.XuX.FittsE. P. (2023). A conditional variational auto-encoder model for reducing musculoskeletal disorder risk during a human-robot collaboration task. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 67(1), 425–431.
15.
SengarS. S.HasanA.BinKumarS.CarrollF. (2024). Generative artificial intelligence: A systematic review and applications. Multimedia Tools and Applications, 84(21), 23661–23700.
16.
ShalyminovI.LeeS.EshghiA.LemonO. (2019). Few-shot dialogue generation without annotated data: A transfer learning approach. arXiv preprint arXiv:1908.05854.
17.
van den BurgG.WilliamsC. (2021). On memorization in probabilistic deep generative models. Advances in Neural Information Processing Systems, 34, 27916–27928.
18.
WenL.YongG.BaekF.LeeS. (2024). Text-to-motion generation for construction worker’s posture coaching applying stochastic ergonomic risk minimization. ASCE International Conference on Computing in Civil Engineering (i3CE), Pittsburgh, PA.
19.
XiaoT.HongJ.MaJ. (2018). Elegant: Exchanging latent encodings with GAN for transferring multiple face attributes. In Proceedings of the European Conference on Computer Vision (ECCV) (pp. 168–184). Springer. https://doi.org/10.1007/978-3-030-01234-2_11
20.
YongG.LiuM.LeeS. (2024a). Automated captioning for ergonomic problem and solution identification in construction using a vision-language model and caption augmentation. Construction Research Congress, 2024, 709–718.
21.
YongG.LiuM.LeeS. (2024b). Explainable image captioning to identify ergonomic problems and solutions for construction workers. Journal of Computing in Civil Engineering, 38(4), 4024022.
22.
ZhaoY.DuC.AbdollahzadehM.PangT.LinM.YanS.CheungN.-M. (2023). Exploring incompatible knowledge transfer in few-shot image generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 7380–7391). IEEE. https://doi.org/10.1109/CVPR56688.2023.00725
23.
ZhuJ.-Y.ParkT.IsolaP.EfrosA. A. (2017). Unpaired image-to-image translation using cycle-consistent adversarial networks. In Proceedings of the IEEE International Conference on Computer Vision (pp. 2223–2232). IEEE. https://doi.org/10.1109/ICCV.2017.244
