We consider predicting the user's head motion in 360 ° videos, with 2 modalities only: past positions and video content (not knowing other users' traces). make two main contributions. First, we re-examine existing deep-learning approaches for this problem identify hidden flaws from a thorough root-cause analysis. Second, results of analysis, design new proposal establishing state-of-the-art performance. re-assessing methods that use both modalities, obtain surprising result they all perform...

We present the CREATTIVE3D dataset of human interaction and navigation at road crossings in virtual reality. The has three main breakthroughs: (1) it is largest motion fully-annotated scenarios (40 hours, 2.6 million poses), (2) captured dynamic 3D scenes with multivariate – gaze, physiology, data, (3) investigates impact simulated low-vision conditions using eye tracking under real walking conditions. Extensive effort been made to ensure transparency, usability, reproducibility study...

We study replication mechanisms that include Reed-Solomon type codes as well network coding in order to improve the probability of successful delivery within a given time limit. propose an analytical approach compute these and effect on performance while optimizing parameters govern routing.

Streaming Virtual Reality (VR), even under the mere form of 360° videos, is much more complex than for regular videos because to lower required rates, transmission decisions must take user's head position into account. The way user exploits her/his freedom therefore crucial network load. In turn, moves depends on video content itself. VR however a whole new medium, which film-making language does not exist yet, its "grammar" only being invented. We present strongly inter-disciplinary...

While Virtual Reality (VR) represents a revolution in the user experience, current VR systems are flawed on different aspects. The difficulty to focus naturally headsets incurs visual discomfort and cognitive overload, while high-end require tethered powerful hardware for scene synthesis. One of major solutions envisioned address these problems is foveated rendering. We consider problem streaming stored 360° videos headset equipped with eye-tracking rendering capabilities. Our end research...

In this paper, a new class of low-density parity-check (LDPC) codes, named hybrid LDPC is introduced. Hybrid codes are characterized by an irregular connectivity profile and heterogeneous orders the symbols in codeword. It shown particular that can be asymptotically optimized using density evolution (DE) framework, technique to maximize minimum distance code presented. Numerical assessment performances provided, comparing them protograph-based multiedge-type (MET) codes. allow achieve...

The streaming transmissions of 360° videos is a major challenge for the development Virtual Reality, and require reliable head motion predictor to identify which region sphere send in high quality save data rate. Different predictors have been proposed recently. Some these works similar evaluation metrics or even share same dataset, however, none them compare with each other. In this article we introduce an open software that enables evaluate heterogeneous prediction methods on various...

From a user perspective, immersive content can elicit more intense emotions than flat-screen presentations. system efficient storage and distribution remain challenging, must consider attention. Understanding the connection between attention, is therefore key. In this article, we present new dataset, PEM360 of head movements gaze recordings in 360° videos, along with self-reported emotional ratings valence arousal, continuous physiological measurement electrodermal activity heart rate. The...

Prediction of head movements in immersive media is key to design efficient streaming systems able focus the bandwidth budget on visible areas content. Numerous proposals have therefore been made recent years predict 360° images and videos. However, performance these models limited by a main characteristic motion data: its intrinsic uncertainty. In this article, we present an approach generate multiple plausible futures videos, given common past trajectory. Our method provides likelihood...

In Delay Tolerant Networks (DTNs) the core challenge is to cope with lack of persistent connectivity and yet be able deliver messages from source destination. particular, routing schemes that leverage relays' memory mobility are a customary solution in order improve message delivery delay. When large files need transferred destination, not all packets may available at prior first transmission. This motivates us study general packet arrivals source, derive performance analysis...

This paper tackles the issue of reliable transport in delay-tolerant mobile ad hoc networks (DTNs) that are operated by some opportunistic routing algorithm. We propose a mechanism relies on acknowledgements (ACKs) and coding at source. The various versions problem depending buffer management policies formulated fluid model based mean-field approximation is derived for designed mechanism. allows both mean file completion time energy consumption to be expressed up delivery last ACK accuracy...

Virtual Reality (VR) technology enables "embodied interactions" in realistic environments where users can freely move and interact, with deep physical emotional states. However, a comprehensive understanding of the embodied user experience is currently limited by extent to which one make relevant observations, accuracy at observations be interpreted.

The surge of video traffic is a challenge for service providers that need to maximize Quality Experience (QoE) while optimizing the cost their infrastructure. In this paper, we address problem routing multiple HTTP-based Adaptive Streaming (HAS) sessions QoE. We first design QoS-QoE model incorporating different QoE metrics which able learn online network variations and predict impact on representative classes adaptation logic, motion client resolution. Different are then combined into score...

A class of two-bit message passing decoders for decoding column-weight-four LDPC codes over the binary symmetric channel is proposed. The thresholds various in this are derived using density evolution. For a specific decoder, sufficient conditions correcting all error patterns with up to three errors derived.

Adaptive bitrate (ABR) algorithms are used in streaming media to adjust video or audio quality based on the viewer's network conditions provide a smooth playback experience. With rise of virtual reality (VR) headsets, 360° is growing rapidly and requires efficient ABR strategies also adapt user's head position. However, research this field often difficult compare due lack reproducible simulations. To address problem, we SMART360, simulation environment motion prediction adaptive bitrates...

First unequal error protection (UEP) proposals date back to the 1960's (Masnick and Wolf; 1967), but now with introduction of scalable video, UEP develops a key concept for transport multimedia data. The paper presents an overview some new approaches realizing properties in physical transport, especially multicarrier modulation, or LDPC Turbo codes. For bit-loading together hierarchical modulation is described allowing arbitrary number classes, SNR margins between bits per class. In coding,...

In this paper, we propose to study and optimize a very general class of LDPC codes whose variable nodes belong finite sets with different orders. We named hybrid codes. Although efficient optimization techniques exist for binary more recently non-binary LDFC codes, they both exhibit drawbacks due reasons. Our goal is capitalize on the advantages families by building (or small set order) parts their factor graph representation. The obviously larger than existing types which gives degrees...

Head motion prediction is an important problem with 360° videos, in particular to inform the streaming decisions. Various methods tackling this deep neural networks have been proposed recently. In article, we introduce a new architecture, named TRACK, that benefits both from history of past positions and knowledge video content. We show TRACK achieves state-of-the-art performance when compared against all recent approaches considering same datasets wider horizons: 0 5 seconds.

Virtual reality (VR) offers extraordinary opportunities in user behavior research to study and observe how people interact immersive 3D environments. A major challenge of designing these experiences tasks, however, lies bridging the inter-relational gaps perception between designer, user, scene. Paul Dourish identified three perception: ontology scene representation designer interpretation, intersubjectivity task communication intentionality user's intentions designer's interpretations. We...

Cinematic Virtual Reality (VR) has the potential of touching masses with new exciting experiences, but faces two main hurdles: one is ability to stream these videos, another their design and creation. Indeed, rates are much higher in addition discomfort sickness that might arise fully immersive experience a headset, users get lost when exploring 360° videos miss elements required understand underlying plot. We take an innovative approach by addressing jointly creation streaming problems....

Despite exciting prospects, the development of 360° videos is persistently hindered by difficulty to stream them. To reduce data rate, existing streaming strategies adapt video rate user's Field View (FoV), but predicting FoV and persistent lack bandwidth are important obstacles achieve best experience. In this article we exploit recent findings on human attention in VR introduce a new additional degree freedom for algorithm leverage: Virtuall Walls (VWs) designed translate limitation into...