cd /news/computer-vision/what-i-learned-about-3d-reconstructi… · home topics computer-vision article
[ARTICLE · art-41236] src=dev.to ↗ pub= topic=computer-vision verified=true sentiment=↑ positive

What I Learned About 3D Reconstruction in Week 1 of My AI Internship at PreserveMy.World

An intern at TechRealm x PreserveMy.World compared five 3D reconstruction methods for preserving cultural heritage sites, including COLMAP, NeRF, Gaussian Splatting, monocular depth estimation, and multi-view stereo. The intern wrote a Python script simulating monocular depth estimation and generated a depth map, confirming the pipeline concept. The nonprofit uses AI to document historical landmarks like Lahore's Badshahi Masjid.

read1 min views1 publishedJun 26, 2026

This week marked the beginning of my internship at TechRealm x PreserveMy.World —

a nonprofit using AI to preserve cultural heritage sites in 3D. I'm on the Core Tech

track, focused on Platform & Web Engineering.

PreserveMy.World is a nonprofit initiative that uses AI and 3D reconstruction to

document and preserve historical sites — starting with Lahore's heritage landmarks

like Badshahi Masjid.

I compared five methods for reconstructing heritage sites in 3D:

1. COLMAP (Structure-from-Motion) Takes multiple photos and produces a 3D point cloud. Great starting point for PMW.

2. NeRF (Neural Radiance Fields) Creates photorealistic 3D scenes from images. Needs a GPU but produces stunning results.

3. Gaussian Splatting

Fast, high-quality 3D rendering — ideal for web platforms like PreserveMy.World.

4. Monocular Depth Estimation

Works from a single image or video. Most practical for field use with limited equipment.

5. Multi-View Stereo (MVS) Dense reconstruction from multiple calibrated images. Good for building facades.

I wrote a Python script simulating monocular depth estimation and generated a depth

map output. Small experiment, but it confirmed the pipeline concept works.

The first attempt at down a heritage image in Python returned a 403 error —

so I switched to a simulated depth map instead. Real-world data acquisition will need

proper API access or local images.

I'll continue researching 3D reconstruction methods, run more experiments,

This post is part of my TechRealm x PreserveMy.World internship journey. GitHub: https://github.com/Anfey-SE/PMW-day1

── more in #computer-vision 4 stories · sorted by recency
── more on @techrealm 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/what-i-learned-about…] indexed:0 read:1min 2026-06-26 ·