{"id":105302,"date":"2022-10-13T15:36:48","date_gmt":"2022-10-13T13:36:48","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/otkryt-kod-sistemy-mashinnogo-obucheniya-dlya-generaczii-realistichnyh-dvizhenij-cheloveka"},"modified":"2022-10-13T15:36:48","modified_gmt":"2022-10-13T13:36:48","slug":"otkryt-kod-sistemy-mashinnogo-obucheniya-dlya-generaczii-realistichnyh-dvizhenij-cheloveka","status":"publish","type":"post","link":"https:\/\/prohoster.info\/en\/blog\/news\/otkryt-kod-sistemy-mashinnogo-obucheniya-dlya-generaczii-realistichnyh-dvizhenij-cheloveka","title":{"rendered":"The code for a machine learning system that generates realistic human movements has been made available","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
A group of researchers from Tel Aviv University has released the source codes related to the Motion Diffusion Model (MDM), a machine learning system that enables the generation of realistic human movements. The code is written in Python using the PyTorch framework and is distributed under the MIT license. For experimentation, both pre-trained models and the option to train models from scratch using provided scripts are available, for example, utilizing the HumanML3D collection of 3D human images. A GPU with CUDA support is required for system training. <\/p>\n
The application of traditional methods for animating human movements is complicated by challenges stemming from the vast variety of possible movements and the difficulty in formally describing them, as well as the high sensitivity of human perception to unnatural movements. Previous attempts to use generative machine learning models faced issues with quality and limited expressiveness. <\/p>\n
In the proposed system, an attempt has been made to use diffusion models for generating movements, which are inherently better suited for simulating human motions but are not without drawbacks, such as high computational resource requirements and management complexity. To minimize the disadvantages of diffusion models, a neural network with a 'transformer' architecture is employed in MDM, utilizing sample prediction instead of noise prediction at each stage, simplifying the prevention of anomalies such as losing contact of the foot with the surface. <\/p>\n
To control generation, there is the option to apply a natural language description of the action (for example, 'a person walks forward and bends down to pick something up from the ground') or to use standard actions like 'running' and 'jumping'. The system can also be used for editing movements and filling in lost details. Researchers conducted testing in which participants were asked to choose the higher-quality result from several options \u2014 in 42% of cases, people preferred the synthesized movements over the real ones.
![\"Play](\"https:\/\/img.youtube.com\/vi\/rVkIDj5wgjs\/hqdefault.jpg\")
\r\n