About InterDigital
InterDigital is a global research and development company focused primarily on wireless, video, artificial intelligence (“AI”), and related technologies. We design and develop foundational technologies that enable connected, immersive experiences in a broad range of communications and entertainment products and services. We license our innovations worldwide to companies providing such products and services, including makers of wireless communications devices, consumer electronics, IoT devices, cars and other motor vehicles, and providers of cloud-based services such as video streaming. As a leader in wireless technology, our engineers have designed and developed a wide range of innovations that are used in wireless products and networks, from the earliest digital cellular systems to 5G and today’s most advanced Wi-Fi technologies. We are also a leader in video processing and video encoding/decoding technology, with a significant AI research effort that intersects with both wireless and video technologies.
Founded in 1972, InterDigital is listed on Nasdaq.
InterDigital is a registered trademark of InterDigital, Inc.
For more information, visit: www.interdigital.com .
Summary
Context
Neural Video Codecs (NVCs) use learned representations and neural-network-based transforms to compress video. Recent codecs such as Microsoft’s DCVC-RT [1] demonstrate that neural video compression is approaching the performance required for real-time applications.
These codecs differ from conventional video codecs in their compressed representations, prediction dependencies, model requirements, and decoder state. Their bitstreams therefore do not naturally fit existing media containers and packetization formats.
Media over QUIC Transport (MOQT) [2] is being developed within the IETF to support low-latency media distribution over QUIC. The NMSF Internet-Draft [3] proposes an initial method for packaging neural video frames as MoQ Groups and Objects. This proposal requires implementation and experimental evaluation against realistic codec behavior and network conditions.
Objective
The internship will investigate how an existing real-time neural video codec, such as Microsoft DCVC-RT, can be packaged and transported over MOQT. The student will study the state of the art, implement and evaluate an NVC proposal, identify its limitations, and propose improvements or an alternative packetization approach where justified.
Responsibilities
Review the state of the art in neural video compression, real-time media transport, QUIC, MOQT, and media packetization.
Analyze NVC bitstreams, frame dependencies, decoder state, and transport requirements.
Implement NMSF-based packetization and integrate the codec with an MOQT sender and receiver.
Develop a reproducible testbed and evaluate latency, overhead, robustness, recovery behavior, video quality, and computational performance.
Analyze the experimental results and propose improvements to NMSF or an alternative packetization design.
Qualifications
Master’s student in computer science, telecommunications, signal processing, or a related field.
Knowledge of computer networks, multimedia systems, video coding, machine learning.
Programming experience, preferably in Python and C or C++.
Ability to read and analyze technical standards and research publications.
Familiarity with QUIC, RTP, video compression, video-quality evaluation, neural compression would be an advantage.
Keywords
Neural video coding, NVC, DCVC-RT, Media over QUIC, MOQT, NMSF, QUIC, real-time media, video packetization, low-latency streaming.
Expected Outcomes
A state-of-the-art analysis of NVC transport over QUIC and MOQT.
A functional NVC-over-MOQT evaluation testbed.
An implementation and experimental evaluation of NMSF.
Recommendations for improving NMSF or an experimentally supported alternative.
A Master’s thesis and technical presentation.
Depending on the results, a potential standards contribution or Internet-Draft and an…