Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | https://arxiv.org/abs/2509.14290 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866911160561827840 |
|---|---|
| author | Kadijani, Niusha Sabri Manjunath, Yoga Suhas Kuruba Bi, Xiaodan Zhao, Lian |
| author_facet | Kadijani, Niusha Sabri Manjunath, Yoga Suhas Kuruba Bi, Xiaodan Zhao, Lian |
| contents | We propose QLook, a quantum-driven predictive framework to improve viewport prediction accuracy in immersive virtual reality (VR) environments. The framework utilizes quantum neural networks (QNNs) to model the user movement data, which has multiple interdependent dimensions and is collected in six-degree-of-freedom (6DoF) VR settings. QNN leverages superposition and entanglement to encode and process complex correlations among high-dimensional user positional data. The proposed solution features a cascaded hybrid architecture that integrates classical neural networks with variational quantum circuits (VQCs)-enhanced quantum long short-term memory (QLSTM) networks. We utilize identity block initialization to mitigate training challenges commonly associated with VQCs, particularly those encountered as barren plateaus. Empirical evaluation of QLook demonstrates a 37.4% reduction in mean squared error (MSE) compared to state-of-the-art (SoTA), showcasing superior viewport prediction. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_14290 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | QLook:Quantum-Driven Viewport Prediction for Virtual Reality Kadijani, Niusha Sabri Manjunath, Yoga Suhas Kuruba Bi, Xiaodan Zhao, Lian Quantum Physics Systems and Control We propose QLook, a quantum-driven predictive framework to improve viewport prediction accuracy in immersive virtual reality (VR) environments. The framework utilizes quantum neural networks (QNNs) to model the user movement data, which has multiple interdependent dimensions and is collected in six-degree-of-freedom (6DoF) VR settings. QNN leverages superposition and entanglement to encode and process complex correlations among high-dimensional user positional data. The proposed solution features a cascaded hybrid architecture that integrates classical neural networks with variational quantum circuits (VQCs)-enhanced quantum long short-term memory (QLSTM) networks. We utilize identity block initialization to mitigate training challenges commonly associated with VQCs, particularly those encountered as barren plateaus. Empirical evaluation of QLook demonstrates a 37.4% reduction in mean squared error (MSE) compared to state-of-the-art (SoTA), showcasing superior viewport prediction. |
| title | QLook:Quantum-Driven Viewport Prediction for Virtual Reality |
| topic | Quantum Physics Systems and Control |
| url | https://arxiv.org/abs/2509.14290 |