Saratoga, CA / April 9, 2018 – MulticoreWare has formed the Media and AI-analytics business unit to leverage its expertise in GPU computing, deep-learning, and video processing to solve problems faced by the media industry at large. This business unit will deliver solutions in video codecs, real time object detection and tracking, object counting, pose estimation and identification that may be used by several vendors in broadcasting, OTT-streaming, and surveillance & security.
In the video processing space, MulticoreWare’s offerings include x265, the most popular open-source HEVC encoder, UHDkit, a software encoder that enables commercial use-cases of existing open-source AVC and HEVC encoders, and UHDcode, a software HEVC decoder that enables using HEVC on a variety of end-user devices. MulticoreWare has recently accelerated x265 with Machine Learning and CNN techniques.
LipSync is an AI based audio-video sync detection library that detects synchronization issues between the audio and video streams of any content without having access to the source, deployable in both real-time and VoD work flows. To solve this complex problem plaguing the media industry today, LipSync uses CNNs and our expertise in media processing to detect faces, track lips, and analyze human speech. TextSync leverages AI solutions to solve the synchronization problem for audio, and subtitles.
MulticoreWare’s additional solutions in the Media and AI-analytics space enable multi-person tracking with unique IDs through the use of face, age, and gender recognition, and pose-estimation through multi-keypoint identification to detect complex poses even in the presence of occlusion. These solutions operate at real-time speeds on embedded devices including DSPs, and FPGAs making them relevant to use as building blocks for applications in several verticals including security, surveillance, gesture recognition, and general optimization of personnel productivity in enterprise environments.
“Working with MulticoreWare has allowed MagikVision to quickly incorporate advanced machine-learning techniques into our sports video autoproduction system. MulticoreWare’s new pose detection and identification technologies will be featured in our upcoming products” says Spero Koulouras, CEO of MagikVision.
“We are uniquely positioned to deliver state-of-the-art high performance AI based smart analytics solutions for any media content with our core competencies in Video processing, Deep Learning, and Micro architecture aware performance optimization. LipSync is a classic example of using all the above skills to solve an important problem for the Media industry” says Arun Ramanathan, Vice President & GM, Media & AI Analytics Business Unit of MulticoreWare.
MulticoreWare is demonstrating its Media and AI Analytics products including HEVC Codecs, ML & CNN based x265 encoders, LipSync, Text Sync, Pose detection, Object detection and Tracking products at MulticoreWare’s booth at NAB 2018 and at ISC WEST 2018 in Las Vegas.