Expected Outcome

What we expect

Expected Test Results

We incorporate Peak Signal-to-Noise Ratio (PSNR), a metric in image quality evaluation. PSNR is defined as:

Where:

• L represents the maximum possible pixel value.

• MSE is the Mean Squared Error, computed as the average squared difference between corresponding pixels in the reference image and the MMRF-generated image.

This quantifies the accuracy and fidelity of MMRF's output by comparing it to reference data. Achieving a higher PSNR signifies that MMRF generates multiview images with minimal noise and loss of image quality, contributing to evaluation of performance in rendering and scene generation.

We expect MMRF to outperform NeRF in rendering volumetric scenes with higher structural similarity index and more optimal FEA results as it incorporates more diverse multimodal inputs such as audio and video footage.

Furthermore, we posit that unlike 3D-GPT, Infinigen and NeRF which focus on asset rendering, we will generate a prompt-based physical-visual environment with higher frame rate stability and reduced latency measurements. This is via intricate semantic mapping that customises physics with real-time tensor operations.

In civil engineering, AI-driven VR empowers students to interact with virtual structures, exploring stress and strain dynamics [8]. This tool offers exposure to diverse architectural models, connecting theory and real-world practice, transforming traditional learning [9].

Impact of Research

This adoption is poised to enhance academic outcomes and classroom engagement. VR simulations could spawn monetary efficiencies for education institutions. We foresee this paradigm shift fostering elevated employment prospects for graduates and propelling seminal research on GAI and VR’s integration in pedagogy.

The technology can be further extended into humanities education. For the study of geography, users can construct VR worlds to visualise tectonic plates, coastal waves, tropical forests, rock types and formation. Users can customise the physics to witness varying levels of fractures, folds, and faults on natural landforms. They can virtually build any landforms with simple text prompts to instantly generate photorealistic results [10].