Milestone

Abstract: In the era of Industry 4.0, manufacturing and production systems were revolutionized by increasing operational efficiency and developing and implementing new business models, services, and products. Concretely, the milestone set for Industry 4.0 was to improve the sustainability and efficiency of production systems. By extension, the emphasis was focused on both the digitization and the digitalization of systems, providing room for further improvement. However, the current technological evolution is more system/machine-oriented, rather than human-oriented. Thus, several countries have begun orchestrating initiatives towards the design and development of the human-centric aspect of technologies, systems, and services, which has been coined as Industry 5.0. The impact of Industry 5.0 will extend to societal transformation, which eventually leads to the generation of a new society, the Society 5.0. The developments will be focused on the social and human-centric aspect of the tools and technologies introduced under the framework of Industry 4.0. Therefore, sustainability and human well-being will be at the heart of what comes next, the Industry 5.0, as a subset of Society 5.0. Industry 5.0 will build on the foundations laid during Industry 4.0 by emphasizing human-centered, resilient, and sustainable design. Consequently, the authors in this research work, through a critical literature review, aim to provide adequate reasoning for considering Industry 5.0 as a framework for enabling the coexistence of industry and emerging societal trends and needs. The contribution of this research work extends to the provision of a framework to facilitate the transition from Industry 4.0 to Society 5.0.

Abstract: Generating temporally coherent high fidelity video is an important milestone in generative modeling research. We make progress towards this milestone by proposing a diffusion model for video generation that shows very promising initial results. Our model is a natural extension of the standard image diffusion architecture, and it enables jointly training from image and video data, which we find to reduce the variance of minibatch gradients and speed up optimization. To generate long and higher resolution videos we introduce a new conditional sampling technique for spatial and temporal video extension that performs better than previously proposed methods. We present the first results on a large text-conditioned video generation task, as well as state-of-the-art results on established benchmarks for video prediction and unconditional video generation. Supplementary material is available at https://video-diffusion.github.io/