Comparative Evaluation of 3D Reconstruction Methods for Immersive Visualization of Laboratory Objects

Noosaga Placements

• Neural Renderingframework90%
  The paper evaluates NeRF and Gaussian splatting, which are neural rendering techniques, and uses them to generate holographic models.
  we compared four approaches: photogrammetry, a neural radiance field (NeRF)–based method, Gaussian splatting, and LiDARGaussian splatting renders each of the millions of captured points in an object as a Gaussian distribution of characteristics, including position, color, size, and opacity, which are then projected onto a screen
• Graphics Renderingsubfield90%
  The paper focuses on 3D reconstruction and rendering methods (NeRF, Gaussian splatting) for creating holographic models, which are core topics in graphics rendering.
  we compared four approaches: photogrammetry, a neural radiance field (NeRF)–based method, Gaussian splatting, and LiDARGaussian splatting renders each of the millions of captured points in an object as a Gaussian distribution of characteristics, including position, color, size, and opacity, which are then projected onto a screen
• Inverse Rendering and Scene Reconstructionframework85%
  The paper evaluates 3D reconstruction methods including NeRF and photogrammetry, which are inverse rendering and scene reconstruction techniques.
  we compared four approaches: photogrammetry, a neural radiance field (NeRF)–based method, Gaussian splatting, and LiDARA radiance field maps the interaction of light, color, and density with individual elements of an object as a function of position and viewing angle, and the object representation is then constructed through neural network training
• Computational Imagingsubfield85%
  The study evaluates computational imaging techniques for 3D reconstruction, including NeRF and photogrammetry, which are key computational imaging methods.
  A radiance field maps the interaction of light, color, and density with individual elements of an object as a function of position and viewing angle, and the object representation is then constructed through neural network trainingPhotogrammetry creates 3D structures from a series of approximately 200 two-dimensional photographs taken of an object from different angles and heights
• Geometric and Physical Reconstructionframework80%
  The paper compares methods for geometric and physical reconstruction of 3D objects, including photogrammetry and LiDAR.
  Photogrammetry creates 3D structures from a series of approximately 200 two-dimensional photographs taken of an object from different angles and heightsLiDAR emits laser pulses toward an object and uses the return time to calculate distances. The resulting point cloud is a direct measure of object geometry
• Computer Visionsubfield70%
  The paper involves 3D reconstruction from images, a core computer vision task, and uses NeRF, a neural network-based method.
  we compared four approaches: photogrammetry, a neural radiance field (NeRF)–based method, Gaussian splatting, and LiDARNeural radiance fields were first described in 2021 and, like the other methods used here, require scene capture from multiple positions
• Deep Learningsubfield60%
  NeRF is a deep learning-based method, and the paper evaluates its performance, but the study is not primarily about deep learning techniques.
  A radiance field maps the interaction of light, color, and density with individual elements of an object as a function of position and viewing angle, and the object representation is then constructed through neural network training

Abstract

In this study, we examined whether current 3D reconstruction methods can support the creation of realistic holographic representations of laboratory objects for educational use. In this regard, we compared four approaches: photogrammetry, a neural radiance field (NeRF)-based method, Gaussian splatting, and LiDAR. These methods were used to generate holographic models of common laboratory items and their fidelity was evaluated by graduate students. Participants assessed the models for shape, color, texture, and visual defects using a repeated-measures design. Across objects, the NeRF-based method produced the most consistently high-fidelity representations, particularly for transparent, reflective, or low-texture items that were difficult to capture with other approaches. Shape and color were generally reproduced more successfully than texture, suggesting that some visual properties remain more challenging to represent accurately in educational holograms. Beyond identifying the strengths and limitations of each reconstruction method, the study demonstrates a practical workflow for creating immersive learning objects that may support pre-laboratory preparation, spatial reasoning, and student engagement in AR/MR-based educational environments. These findings offer design-relevant insights for educators and researchers developing immersive digital learning experiences.

Paper Context