Source: Deep Learning on Medium
SKY ENGINE invited by NVIDIA to lead the workshops at GPU Technology Conference in Silicon Valley, US in March 2020
Sky Engine has been invited second year in a row to set up the Deep Learning Workshops providing Instructor-Led Training at the GPU Technology Conference (GTC), taking place March 22–26, 2020 at the San Jose McEnery Convention Center in San Jose, California.
Sky Engine will provide 2 hours long training session “Raytracing for Deep Neural Networks training” including basics of generative 3D scene preparation for AI-driven computer vision training in virtual reality with procedural geometry, lighting and texturing, as well as creation of auto-balanced rendering data sources ready for deep learning. Everything available during hands-on session performed by Sky Engine’s experts. Sky Engine AI was leading the Workshops also during Nvidia’s GTC 2019 having a full house of 150 people.
Sky Engine will showcase full-stack evolutionary artificial intelligence (AI) engine providing automatic data generation and deep learning pipeline for computer vision applications. Sky Engine can generate data using proprietary, dedicated rendering system where images come already labeled being ready for the deep learning. Rendering system is tightly coupled with the deep learning pipeline to ensure the evolution of Sky Engine — during the training Sky Engine can spot ambiguous situations that deteriorate the accuracy of the AI model and asks renderer to provide more imagery data to reflect problematic situations that the deep learning could improve. Also, Sky Engine improves with every conducted experiment.
Sky Engine is planned to be available on the market around mid 2020 in a Software-as-a-Service (SaaS) model with the pricing strategy based on the value it will bring to the customers in their applications of computer vision.
Sky Engine combines a ray tracing rendering system that is directly integrated with the AI framework and designed to render images for training machine/computer vision AI in virtual environments. Sky Engine generates training data using virtual computer generated imagery (CGI) scenes. By automatically changing parameters like camera position, object size and location, material properties and lighting in the CGI scene, Sky Engine can generate a massive number of already annotated images for AI computer vision training ensuring high accuracy of the inference.