Research Nikos discussed on The TWIML AI Podcast

Automatic TRANSCRIPT

So so commercial GPO and we can get almost deal time as speed. Or call. It to maybe start to wrap things up at love for each you to kind of put this paper and maybe talk about the things that your either your top lessons learned are the thing that you excited about with this paper and his contributions and kind of how you see it. Moving forward in your own Research Nikos, you want to start us off the yeah. Can Start Show I doing M, B I in September under the I didn't know a lot of things for three D. human mode links show. These paper was is, is a very, very good the first experience and I think the most important one of the most important features of the paper is that it is clear i. mean the contributions are clear we use. A lot of an pair three data to refine to the predictions actually to predict to the predictions. Actually. Three. Meshes, that were predicted mostly from the points and. our implementation is converges foss. Gun. create high-quality in state of the art Date a The predictions are a quite high quality and the state of the art. And also one important thing is that the we've made all the goal to end the data obliquely. We have detailed instructions for everyone. So it's really easy for anyone if you want or C- wants to run and use our model and methods, we have clear instructions to do it right Muhammad. So it seems that Vibe is working quite on the for Volition V. have but in real life or in into wild videos, a little difficult situations end. Motion is quite A. Quite difficult to model because she wants capable of lots of different kinds of moments and also in real life situations we have. Like objects and scenes on. Voss. We have all skulls by. Again objects or other people. These kind of problems are still waiting to be tackled with and..

Coming up next