5 offres trouvées

Summary In this internship at the London AI Video Lab, the objective is to design computationally efficient video decoders in an AI-based video compression codec. Current AI-based video compression models outperform conventional codecs, like HEVC, VVC and AV1. However, this comes at the cost of impractical compute requirements: at decode, current AI-based video compression decoders are several orders of magnitude more complex than conventional video compression decoders. The goal of the internship is to design efficient AI-based decoders that leverage spatial sparsity to reduce their computational complexity. This work will be seen as one step forward toward the deployment of end-to-end trained AI-based video compression models. The goal will be to find and review potential existing methods of spatial sparsity in AI-based video models. In a second step, spatially sparse AI-based decoders will be designed, implemented and integrated into the London AI Video Lab's end-to-end...

Summary In this internship at the London AI Video Lab, the objective is to study fixed-point arithmetic solutions for ensuring bit-exact video compression in AI-based video codecs. Current AI-based video compression models outperform conventional codecs, like HEVC, VVC and AV1. However, AI-based video compression models are trained using floating-point arithmetic. Unfortunately, floating point arithmetic is insufficient to ensure bit-exact execution. Bit-exact execution is needed to ensure encoded bitstreams are universally decodable across any device. Fixed-point arithmetic is a potential solution to this problem. The goal of the internship is to determine a fixed-point arithmetic setup capable of ensuring bit-exactness while maintaining model performance. This work will be seen as one step forward toward the deployment of end-to-end trained AI-based video compression models. The goal will be to study various fixed-point arithmetic setups for layers and components of AI-based...

Summary The InterDigital Immersive Video Lab is focused on advancing avatar-based communication. In this context, accurate and detailed human 3D data is essential for reliably estimating features such as facial expressions, body posture, skin texture, lighting conditions, background segmentation, hair segmentation, and more. InterDigital has developed a proprietary framework for generating synthetic 3D facial image datasets. The next step is to enhance this framework to support the creation of 4D video sequences featuring full-body characters. This internship will contribute to the development and refinement of this extended capability. Responsibilities The goal of the internship will be to: Study state of the art on face & body analysis models and synthetic data generation Update the framework to full body with body animation Upgrade the framework to output video sequences Create and/or set up avatar elements (clothes, accessories, jewels etc.) Generate video...

Summary Research-oriented internship on multi-user Quality of Experience (QoE) for XR application, the intern will survey the state of the art, identify and formalize QoE metrics specific to multi-user XR, and propose/enhance a composite QoE metric with corresponding instrumentation. The intern will design and run controlled experiments to collect metrics across concurrent users, analyze correlations with perceived experience, and iterate on the model. The work may culminate in planning and conducting user studies in the Rennes lab. Expected outcomes: a literature review, a measurement framework/prototype, practical recommendations for QoE-aware adaptation in multi-user XR, participation in a scientific paper, and potential disclosures. Responsibilities State of the art in QoE in XR, and Multi-User scenarios Propose or refine a composite QoE metric and map each factor to measurable KPIs and thresholds Design and implement instrumentation to capture interaction signals across...

Summary Multi-Modal Large Language Models (MLLMs) are increasingly capable of processing and reasoning over diverse input modalities such as text, images, audio, and video. However, running such models in real-time or resource-constrained environments poses significant challenges in terms of bandwidth and compute requirements. Distributing the processing of models between client and server (a "distributed computing" approach) is a promising solution. While traditional distributed inferencing has been applied successfully to DNN models, extending this paradigm to MLLMs is a novel and impactful use case. This internship aims to demonstrate the feasibility of distributed MLLMs inference approach, where MLLM components are distributed across two endpoints, coordinated through an agentic orchestration. Responsibilities The internship will be involved in the following tasks: Survey the recent advances in MLLMs, Select representative models, and a representative agentic...