🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Congratulations to Metin Bicer on the first published work from his PhD: Altering the strength of the muscles crossing the lower limb joints only affects knee joint reaction forces doi.org/10.1016/j.gait… Luca Modenese Andrew Phillips

Metin Bicer will present a new augmentation method we developed for mocap datasets. The trained network can generate infinite distinct, yet realistic, walking data for further use! 🤯🤯 #GANs #DeepLearning WCB 2022 Andrew Phillips

Congratulations to Metin Bicer on his latest Journal of Biomechanics publication on using generative deep learning to augment gait lab datasets Luca Modenese Andrew Phillips Alessandro Melis doi.org/10.1016/j.jbio…

We have just published a new data augmentation method for biomechanical datasets based on deep generative learning in Journal of Biomechanics! Paper link: bit.ly/3C9o5wy #GANs #DeepLearning 1/8

Using adversarial autoencoders, Metin Bicer generated synthetic 🧪marker trajectories and ground reaction forces, from which we calculated joint angles and moments statistically similar to those of the real dataset used for training. 2/8

It took some time for our models to learn how to walk properly. At the end, the generated walking trials were similar but not identical to the real walking trials. 3/8

We used the generated data to augment an existing real dataset and improve prediction accuracy in a standard deep learning application where inertial measurement unit data were used to predict joint angles and ground reaction forces. 4/8

One of these trials was generated by our models. Can you guess which one is the synthetic one? 5/8

The one on the left was a real trial. What about the following one? 6/8

You can generate, animate (by Plotly) and download your synthetic mocap data (in OpenSim format for now) from the web app (🙌Streamlit) at thisgaitdoesnotexist.streamlitapp.com. 7/8

Huge congratulations to Metin Bicer and co-authors Andrew Phillips Alessandro Melis and Alison McGregor - it's always a great pleasure working with this team! 🎉🥳🎉

Charly Wargnier Streamlit Kofi 🟦 Fanilo Andrianasolo Andrejus Avra Qiusheng Wu JCharis Jesse Lee Foot 🐍📈 Mykola Kozyr 🇺🇦🇩🇪🇪🇺 J. M. Nápoles Sebastian Flores 🌳 Shubham Saboo Shruti Agarwal Yuichiro William J.B. Mattingly Marco Giordano Hugo Le Moine Lisa Hornung discret JP Hwang @[email protected] @MBohunicka Ralf Gabriels @jpcodesss_ Emiliano López Amith Kamath @zBritva Christoph Rieke Mısra Turp Thank you Charly! Kudos deflected to my PhD student Metin Bicer who actually wrote the app! Great work Metin!

😃New article published in J Biomech! "Generative adversarial networks to create synthetic motion capture datasets including subject and gait characteristics", by Bicer et al. 👀sciencedirect.com/science/articl… #journalofbiomechanics