The latest articles on DEV Community by Amna Hafeez (@amna_hafeez). https://dev.to/amna_hafeez https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4004455%2F727b5ef2-1ec8-43ae-8fc2-069192758630.png https://dev.to/amna_hafeez en Amna Hafeez Fri, 26 Jun 2026 19:23:40 +0000 https://dev.to/amna_hafeez/what-i-learned-about-3d-reconstruction-in-week-1-of-my-ai-internship-at-preservemyworld-32fd https://dev.to/amna_hafeez/what-i-learned-about-3d-reconstruction-in-week-1-of-my-ai-internship-at-preservemyworld-32fd <p>This week marked the beginning of my internship at TechRealm x PreserveMy.World — <br> a nonprofit using AI to preserve cultural heritage sites in 3D. I'm on the Core Tech <br> track, focused on Platform &amp; Web Engineering.</p> <h2> What is PreserveMy.World? </h2> <p>PreserveMy.World is a nonprofit initiative that uses AI and 3D reconstruction to <br> document and preserve historical sites — starting with Lahore's heritage landmarks <br> like Badshahi Masjid.</p> <h2> What I Set Up This Week </h2> <ul> <li>Python, Git, GitHub CLI, and VS Code</li> <li>Gemini CLI as my AI coding assistant</li> <li>My first GitHub repo: PMW-day1</li> </ul> <h2> 3D Reconstruction Methods I Researched </h2> <p>I compared five methods for reconstructing heritage sites in 3D:</p> <p><strong>1. COLMAP (Structure-from-Motion)</strong><br> Takes multiple photos and produces a 3D point cloud. Great starting point for PMW.</p> <p><strong>2. NeRF (Neural Radiance Fields)</strong><br> Creates photorealistic 3D scenes from images. Needs a GPU but produces stunning results.</p> <p><strong>3. Gaussian Splatting</strong><br> Fast, high-quality 3D rendering — ideal for web platforms like PreserveMy.World.</p> <p><strong>4. Monocular Depth Estimation</strong><br> Works from a single image or video. Most practical for field use with limited equipment.</p> <p><strong>5. Multi-View Stereo (MVS)</strong><br> Dense reconstruction from multiple calibrated images. Good for building facades.</p> <h2> My Python Experiment </h2> <p>I wrote a Python script simulating monocular depth estimation and generated a depth <br> map output. Small experiment, but it confirmed the pipeline concept works.</p> <h2> What Failed </h2> <p>The first attempt at downloading a heritage image in Python returned a 403 error — <br> so I switched to a simulated depth map instead. Real-world data acquisition will need <br> proper API access or local images.</p> <h2> What's Next </h2> <p>I'll continue researching 3D reconstruction methods, run more experiments, </p> <h2> and contribute to the PreserveMy.World platform as the internship progresses. </h2> <p><em>This post is part of my TechRealm x PreserveMy.World internship journey. <br> GitHub: <a href="https://github.com/Anfey-SE/PMW-day1" rel="noopener noreferrer">https://github.com/Anfey-SE/PMW-day1</a></em></p> python ai machinelearning beginners