Undergraduate Thesis
Ahmedabad University



Title:  Using 3D Scanning for Preservation and Restoration of Indian Sculptural Artefacts


Abstract: This thesis explores the use of accessible 3D scanning techniques to preserve, document and digitally restore Indian sculptural heritage, with a focus on the Sahasra-Bāhu (Sas-Bahu) temples at Nagda, Rajasthan. Sculptural elements from the site were converted into 3D models using photogrammetry. The models were further refined and selectively restored using digital modelling tools, guided by iconographic analysis and textual sources. The project evaluates both the technical feasibility of low-cost 3D documentation and the interpretive potential of digital reconstruction in art-historical study. The study also examines the pedagogical value of interactive 3D visualisations compared to conventional teaching media, using student and professor surveys. 

Keywords: 3D Scanning, Photogrammetry, Digital Heritage, Indian Temple Architecture, Digital Restoration, 3D Modelling, Virtual Heritage, Iconographic Analysis, Pedagogy, Digital Archive

The full thesis can be found at this URL. The viewable 3D models, along with museum labels, can be found at this URL.

Prototype Study


To test the method before officially commencing the thesis, I worked with a team of art-history students to 3D scan some artefacts at a small museum in Ahmedabad called the BJ Institute of Learning in November 2025. We scanned and produced museum-style labels/iconographic analysis for 4 models, which can be found at this URL. This study shaped the overall workflow, from scanning to textual analysis, for the rest of my thesis.

 

Workflow


At the site, I first took 100-150 photos of each sculptural image using a simple phone camera. The images are taken in a circular orbit like fashion to ensure uniform and maximal coverage of the object. After manually filtering out blurry photos, I bring them into RealityScan. The images are first aligned and then turned into a point cloud (a collection of individually coloured vertices that form the overall object shape). At this stage, I crop out some unwanted parts of the geometry to bring the main sculpture into focus. 

  
 
The point cloud is then converted to an untextured polygon model (bottom left image). Next, I UV unwrap the model and generate textures for it (bottom right image).

 

Lastly, I bring the export of the model with a poly-count of approximately 5 million vertices and a single-texture map with 4090 x 4096 resolution into Blender. 

I further removed some parts of the geometry to focus solely on the main figure contained between the pillars, and also placed a simple black texture on the intersections where this geometry was. It’s important to note here that this process of removing excess geometry is interpretative, and it would be perfectly valid to argue that all surrounding sculptural details contribute to the main portions and must not be removed. This can perhaps best be observed in this render of Brahmā-Sāvitrī (left) compared with the site photograph (right) below. 

 


For digital restoration, I go a step further than simple model refining in Blender, as shown above. First, I outlined the portions of the sculpture that were damaged and evaluated the extent of damage. The image of Mahiṣāsuramardinī below was an ideal candidate because most of the damage was restricted to small portions while the rest of the image was still legible. This sculpture had 4 key areas of damage - the face, the trident, the breast region, and the lower-left arm

Next, I modelled basic primitive shapes to match the broken portions with reference to other sculptures at the site.  I then subdivided them to smooth out the shape. I merged these models with the main model using Blender’s remesh feature and used the sculpting tools to smooth out the regions where they intersected with the main model. For the finishing touches, I add some noise-displacement on the restored portions to match the rocky texture of the rest of the geometry. 

   

I was unable to restore the face of this sculpture as I could not sculpt the details correctly, even after several retries. 

An unfortunate consequence of the remesh was that the original UV mapping was lost, and the attached textures also do not map correctly onto the model. Since the restored portions would also require manual retexturing to match the rest, for the scope of the thesis, it was decided to leave this restored model with a simple “matcap” texture that highlights its overall definition.

The image below shows the final restoration result with the matcap texture (left - original geometry, right - digitally restored). 


 


Results


A total of 9 scans were refined and uploaded to the SketchFab archive with annotations and museum labels. A sample screenshot from the archive is shown below.

It is important to note here that the models on SketchFab are not in their original captured resolution. The geometric detail has been reduced to 5-10% of the original to match the website’s strict file size limits and also reduce lag when navigating the model for users. If you are interested in seeing the original scan quality, please reach out to me! 




Projects:








Contact: khevan.thanki@gmail.com                                                                                                                                                                                                                                             LinkedIn