DEV Community: humnaattique4-sys

Pmw final portfolio · MD Humna Attique — PMW TechRealm Internship Portfolio

humnaattique4-sys — Sat, 08 Aug 2026 10:23:56 +0000

About This Page

This page collects my strongest work from the PMW internship: a Structure-from-Motion debugging case study, a deployed interactive AR demo, and the process evidence behind both. It's meant to serve as the final evidence pack for certificate review and as a shareable summary of what I actually built.

Track & Focus

Track: AI-Based 3D Scene Reconstruction / Platform & Web Engineering
Team: Charbagh Collective
Internship: PreserveMy.World × TechRealm — Cultural Heritage Preservation

I worked on 3D reconstruction of Pakistani heritage sites using COLMAP (Structure-from-Motion) and Three.js (interactive AR/visualization), applied to Shalimar Gardens and Lahore Fort, Lahore.

Strongest Artifacts

1. Case Study: Fixing a Failed COLMAP Reconstruction (Shalimar Gardens)

Diagnosed and fixed a failed Structure-from-Motion reconstruction of the Moorcroft Pavilion — improved camera registration from 2/15 to 15/15 and sparse point count from 42 to 842 by curating tightly-overlapping image sets.

Write-up: DEV.to Case Study
Repository: shalimar-gardens-3d-reconstruction
Process evidence: Commit history

2. Lahore Fort — Interactive Three.js AR Demo

Built and deployed an interactive AR viewer using MiDaS-generated depth maps, validated against the COLMAP South Building benchmark dataset.

Live demo: Lahore Fort AR Demo
Repository: Lahore-Fort-3D-Reconstruction
Process evidence: Commit history

3. Additional Technical Writing

What I Learned

The biggest technical lesson from this internship: in Structure-from-Motion pipelines, image overlap density matters more than image quantity. A failed reconstruction isn't always a hardware or software problem — sometimes it's a data curation problem, and the fix is going back to source images with a more deliberate eye for consistent overlap between viewpoints.

Beyond the technical work, I learned to document a debugging process transparently — showing the failed attempt (V1) alongside the fix (V2), rather than only presenting the polished final result. That transparency is what makes evidence credible for review.

Next Steps

Apply the same image-overlap curation approach proactively to future reconstructions, rather than after a failed first attempt
Extend the Lahore Fort AR demo with additional depth-mapped viewpoints
Explore mesh generation from the densified Shalimar Gardens point cloud

PMW × TechRealm Internship — Charbagh Collective — August 2026

Case Study: Fixing a Failed 3D Reconstruction of Shalimar Gardens' Moorcroft Pavilion

humnaattique4-sys — Fri, 07 Aug 2026 13:33:32 +0000

Problem

The first attempt (V1) to reconstruct the Moorcroft Pavilion at Shalimar Gardens using COLMAP Structure-from-Motion nearly failed: only 2 of 15 input cameras were successfully registered, and the sparse point cloud contained just 42 points — far too sparse to represent the structure. The reconstruction was effectively unusable, and the root cause wasn't obvious from the COLMAP logs alone.

My Role

I owned this reconstruction end-to-end for the AI-Based 3D Scene Reconstruction track: sourcing images, running the COLMAP pipeline, diagnosing the failure, and redesigning the input dataset. This was not a group task — the debugging and the fix were my own work.

Process

Diagnosed the failure: reviewed COLMAP's feature-matching and camera-registration output and traced the low registration count to weak image overlap — the V1 image set didn't have enough shared visual features between consecutive shots for COLMAP to triangulate camera poses reliably.
Rebuilt the dataset: went back to Wikimedia Commons and curated a new image set of the Moorcroft Pavilion selected specifically for tight, consistent overlap between adjacent viewpoints, rather than just visual variety or coverage.
Re-ran the pipeline (V2): same COLMAP configuration, new image set.
Verified the result: confirmed all 15/15 cameras registered and checked the resulting sparse point cloud density and coverage against the pavilion's actual geometry.
Documented it: wrote up the full debugging story as a DEV.to technical walkthrough, and updated the project README.

Evidence

Repository: humnaattique4-sys/shalimar-gardens-3d-reconstruction — commit history shows the V1 attempt, the image set swap, and the V2 run
DEV.to walkthrough post documenting the V1→V2 debugging process
COLMAP output logs / sparse reconstruction stats for both runs (camera registration count, point count)

Sparse point cloud output from the V2 COLMAP run (842 points, 15/15 cameras registered):

Result

Camera registration improved from 2/15 to 15/15
Sparse point count improved from 42 to 842
The pipeline went from an unusable reconstruction to a complete, usable one — and the specific cause (image overlap, not camera count or COLMAP settings) is now documented for reuse on the other heritage sites in the project.

What I Learned

Image overlap density, not just image quantity or coverage, is the binding constraint for Structure-from-Motion pipelines like COLMAP. A larger image set with poor overlap performs worse than a smaller, carefully curated one. This also taught me to read COLMAP's intermediate output rather than only judging a run by whether it completed.

Next Improvements

Apply the same overlap-curation approach proactively to future sites before the first COLMAP run
Add a lightweight pre-check to catch weak overlap before running the full pipeline
Extend the V2 dataset with a few additional angles to further densify the point cloud for downstream mesh generation

From 42 Points to 842: Debugging a CPU-Only 3D Reconstruction Pipeline

humnaattique4-sys — Fri, 31 Jul 2026 22:26:28 +0000

Track: AI-Based 3D Scene Reconstruction — PreserveMy.World × TechRealm Internship (Charbagh Collective)

What I built

As part of my Platform & Web Engineering internship with PreserveMy.World, I built a Structure-from-Motion (SfM) pipeline using COLMAP to generate a sparse 3D point cloud of the Moorcroft Pavilion at Shalimar Gardens, Lahore — entirely on a CPU-only Windows machine, with no CUDA GPU available.

What went wrong first (V1)

My first attempt used 43 general images of Shalimar Gardens pulled from Wikimedia Commons. COLMAP could only register 2 out of 43 cameras, producing just 42 sparse 3D points — essentially unusable. The problem: the images covered different parts of a huge garden complex with almost no consistent overlap between them. COLMAP's feature matcher had nothing reliable to connect across images.

What I fixed (V2)

I switched strategy: instead of broad coverage, I curated 15 tightly-overlapping images focused on a single structure, the Moorcroft Pavilion, ensuring 60–80% visual overlap between consecutive shots. That one change took camera registration from 2/43 (about 5%) to 15/15 (100%), and sparse points jumped from 42 to 842 — a 20x improvement.

What AI helped with, and what I verified myself

I used Claude to help debug why registration was failing, to draft documentation summarizing the technical results, and to structure a before/after visual comparison in the repo README. Everything numeric — camera counts, point counts, which images were used — came from my own COLMAP runs; I verified every figure against my actual pipeline output before it went into any documentation. AI didn't run the reconstruction or generate the data — it helped me communicate results I had already produced and verified.

What I'd improve next

Dense reconstruction (patch-match stereo, which produces a full textured mesh rather than a sparse point cloud) requires a CUDA GPU I don't have access to locally. The next step would be running the same pipeline on cloud GPU compute, or capturing a dedicated, tightly-overlapping photo set on-site at Shalimar Gardens rather than relying on scattered public images.

Why this matters for PMW

PreserveMy.World's mission is digitizing cultural heritage sites in Pakistan. Most students and volunteers won't have access to expensive GPU hardware, so proving that a CPU-only sparse reconstruction pipeline can still produce a meaningful, structurally accurate point cloud — as long as the image capture strategy is right — matters for making heritage digitization accessible at scale, not just to teams with lab-grade equipment.

Evidence

Full pipeline, scripts, and commit history: github.com/humnaattique4-sys/shalimar-gardens-3d-reconstruction

Shalimar Gardens, Lahore: A Heritage Overview

humnaattique4-sys — Fri, 10 Jul 2026 21:34:36 +0000

Introduction

There's something a little disarming about the Shalimar Gardens the first time you actually stand in them. You know going in that they're old — nearly four centuries old — and yet the water is still moving, the terraces still step down exactly the way they were meant to, and none of it feels like a museum exhibit behind glass. It feels used. That's really the throughline of this write-up: the gardens as something the Mughals built to solve a real problem, that the British, the Sikhs, and now the Pakistani state have all had to keep solving in their own way ever since. Together with the Lahore Fort, the site holds UNESCO World Heritage status, inscribed in 1981 [1]. What follows traces where the gardens came from, how they were actually built to work, and what it takes to keep them standing today.

Research Methodology

I went into this with one rule for myself, mostly because it's so easy to break: no single travel blog or tourism page gets treated as a final source. There's a lot of recycled, half-sourced writing about Mughal heritage sites online, and it's easy to accidentally launder someone else's guesswork into your own "research." So I built a rough hierarchy instead — official and institutional bodies first, then editorially reviewed cultural or reference organizations, and only after that anything more general, which I used strictly to cross-check a date or a number rather than to cite on its own.

In practice, that meant starting with UNESCO's own World Heritage Centre listing and the Government of Punjab's Walled City of Lahore Authority — both are primary in the sense that they're the actual bodies responsible for the site, not commentators on it. I layered in the Lahore Biennale Foundation and the New World Encyclopedia next, since both are institutionally reviewed rather than crowd-edited. Wikipedia shows up in the references too, but only where its numbers matched what the official sources already said — it was never the deciding source for anything contested.

Every claim below traces back to at least one institutional or editorially reviewed source, cited in-text with a numbered bracket (like this: [1]) rather than a bare hyperlink, with the full reference at the bottom of this post. The source evaluation table further down lays out exactly how each source was weighed before I used it.

Historical Background

Shah Jahan commissioned the gardens in 1641, and construction was finished within about a year [4]. He wasn't inventing the idea from nothing — his father, Jahangir, had already built a Shalimar garden in Kashmir roughly two decades earlier, and the Lahore version was deliberately modeled on it [3]. But Kashmir had the advantage of naturally sloping land, which meant water could cascade downhill on its own. Lahore is flat. So the project needed serious engineering, not just landscaping: Khalilullah Khan and Mulla Alaul Maulk Tuni oversaw the work on site, while Ali Mardan Khan built a canal roughly a hundred miles long just to pull water in from the Kashmir foothills [2]. Without that canal, there's no garden — Lahore simply doesn't have the water table for it otherwise.

The name itself has a surprisingly contested history. One reading traces "Shalimar" to Sanskrit and Kashmiri roots meaning something like "abode of joy" or "abode of streams" [2]. Centuries later, in the Sikh period, a court debate broke out over what the word actually meant — some of Ranjit Singh's courtiers argued it was a Turkic word for pleasure, but Ranjit Singh himself apparently preferred a Punjabi reading that translated closer to "curse of God," and briefly had the gardens renamed Shahla Bagh before the original name stuck again [2]. It's a small detail, but it says something about how many hands and cultures have passed through this one piece of land.

Architectural and Cultural Features

Layout and Terracing

The whole complex is a long walled rectangle, about 658 meters north to south and 258 meters east to west [5]. Rather than one continuous flat garden, it's cut into three descending terraces, each roughly four to five meters below the one before it: Farah Baksh ("bestower of pleasure") at the top, Faiz Baksh ("bestower of goodness") in the middle, and Hayat Baksh ("bestower of life") at the bottom [5]. Walking from top to bottom is meant to feel like a small journey — from a private royal space down toward something more public and communal.

The Water System

This is the part that tends to surprise people. The gardens hold 414 fountains in total, split across the terraces — 105 up top, 152 in the middle, 153 below — all discharging into marble pools called haūz, alongside five cascades, including the well-known Sawan Bhadoon [2]. None of it runs on pumps. The entire system is gravity-fed: water drops from a raised reservoir through a network of channels and sluices, timed and pressured just by the elevation drop between terraces [4]. It's easy to describe the gardens as "beautiful" and skip past the fact that keeping 414 fountains running without a single motor is, honestly, a hydraulics problem before it's an aesthetic one.

Design Philosophy

The underlying idea comes from Persian garden tradition — the Charbagh, or "four-part garden," built around the notion of an earthly paradise where water, plants, and geometry sit in deliberate harmony [2]. It's not just decoration; in Mughal thought, a well-ordered garden was meant to be a physical argument for a well-ordered universe. Every symmetrical path and every timed fountain is quietly making that point.

The Gardens Under Later Rule

The gardens didn't stay frozen in the Mughal era — they kept getting pulled into whatever politics were happening around them. For generations they hosted Mela Chiraghan, a lamp festival held in the gardens, until President Ayub Khan banned it in 1958 [5]. A few years later, in 1962, Ayub Khan's government went further and nationalized the site entirely, partly as a response to the Mian family — who were then closely tied to the gardens — opposing his imposition of martial law [5]. It's a reminder that a heritage site is never really separate from the politics of the moment; ownership and access to a place like this have shifted hands as much for political reasons as for preservation ones.

Preservation Status and Threats

The gardens haven't always been in good shape. UNESCO placed the site on its "World Heritage in Danger" list in 2000, which forced the government's hand, and by 2012 enough restoration work had happened for it to be removed from that list [6]. That doesn't mean the pressure is off. Unregulated development along GT Road keeps encroaching on the site's buffer zone, the Shah Nahar canal that still physically feeds the fountains deals with pollution and encroachment of its own, and decades of weathering and acid rain have visibly worn down the marble and sandstone [6]. As of June 2023, day-to-day management sits with the Punjab Walled Cities and Heritage Areas Authority [7], which is now the body actually responsible for whatever comes next.

Source Evaluation

Here's the credibility breakdown I used while deciding what to trust and how far to trust it:

Source	Type	Publisher	Year Accessed	Credibility Notes
Fort and Shalamar Gardens in Lahore	Primary / Official	UNESCO World Heritage Centre	2026	Highest authority — official inscription record and management data.
Shalimar Gardens, Lahore	Secondary	Wikipedia	2026	Used for cross-checking dates/figures only; not cited alone for contested claims.
Shalimar Garden: Timeless Beauty of Mughal Design	Primary / Official	Walled City of Lahore Authority (Govt. of Punjab)	2026	Government heritage authority directly responsible for the site.
Shalimar Gardens	Secondary	Lahore Biennale Foundation	2026	Established arts/culture institution; editorially reviewed.
Fort and Shalamar Gardens in Lahore	Secondary / Reference	New World Encyclopedia	2026	Peer-reviewed encyclopedic entry citing published academic works.
7 Wonders of Shalimar Garden	Secondary	Bahawalpur.org	2025	Used only for conservation/threat details, corroborated with UNESCO reporting.
History – Shalimar Gardens	Primary / Official	Official Shalimar Gardens site	2026	Site-managed source; used for current administrative details.

Conclusion

What stays with me most from researching this isn't really the beauty of the place — plenty has been written about that — it's how much deliberate problem-solving is hidden underneath it. A flat, hot city was made to hold 414 working fountains with no pumps, using nothing but elevation and timing. Four centuries later, that same site is losing ground to pollution, encroachment, and weathering, one terrace at a time. That's exactly the gap that documentation work — the kind of 3D reconstruction and digital preservation I've been doing through my internship — is meant to close: capturing an accurate, measurable record of a place like this while it's still standing, so that even if the physical fabric keeps eroding, an honest version of it doesn't have to disappear with it.

References

UNESCO World Heritage Centre. "Fort and Shalamar Gardens in Lahore." whc.unesco.org/en/list/171. Accessed July 2026.
Wikipedia contributors. "Shalimar Gardens, Lahore." en.wikipedia.org/wiki/Shalimar_Gardens,_Lahore. Accessed July 2026.
Walled City of Lahore Authority, Government of Punjab. "Shalimar Garden: Timeless Beauty of Mughal Design." walledcitylahore.gop.pk/shalimar-garden. Accessed July 2026.
Lahore Biennale Foundation. "Shalimar Gardens." lahorebiennale.org/lb03-sites/shalimar-gardens. Accessed July 2026.
New World Encyclopedia. "Fort and Shalamar Gardens in Lahore." newworldencyclopedia.org/entry/Fort_and_Shalamar_Gardens_in_Lahore. Accessed July 2026.
Bahawalpur.org. "7 Wonders of Shalimar Garden: History, Architecture & UNESCO Glory." bahawalpur.org/7-wonders-of-shalimar-garden-history-architecture-unesco-glory. 2025.
Shalimar Gardens (official site). "History." shalimargardans.com/history. Accessed July 2026.

What I Learned About 3D Reconstruction — My First Week at PreserveMy.World

humnaattique4-sys — Tue, 23 Jun 2026 21:39:27 +0000

I just started my internship at PreserveMy.World, a project focused on digitally preserving cultural heritage sites using technology. This is my first week and I spent it setting up my environment and diving into 3D reconstruction research.

What is PreserveMy.World?

PreserveMy.World aims to digitally preserve real places — historical buildings, heritage sites, and landmarks — so they can be experienced even if they're damaged or lost. 3D reconstruction is at the heart of this mission.

The 5 Methods I Researched

1. COLMAP / Structure-from-Motion
Takes multiple photos of the same place from different angles and builds a 3D point cloud. Good for reconstructing buildings from tourist photos.

2. NeRF (Neural Radiance Fields)
Uses a neural network to learn how a scene looks from any angle. Produces photorealistic results but needs a strong GPU.

3. Gaussian Splatting
A newer, faster alternative to NeRF. Produces high quality real-time 3D scenes with fewer photos.

4. Monocular Depth Estimation
Estimates depth from just ONE photo. The easiest to get started with — no special equipment needed.

5. Multi-View Stereo
Uses multiple calibrated photos to build a dense 3D mesh. Good for detailed artifact models.

My Experiment — Monocular Depth Estimation on Google Colab

I decided to start with monocular depth estimation since it only needs one image and runs free on Colab.

I used the Depth-Anything-V2 model from HuggingFace to run depth estimation on a synthetic image I created with Python.

The result was really cool — the model correctly identified which parts of the image were closer and which were further away, shown through a plasma color map (yellow = close, purple = far).

What Failed

My first two attempts failed because I tried downloading images from Wikipedia URLs which got blocked. I fixed it by generating a synthetic image using Python's PIL library instead.

What I Learned

Monocular depth estimation is a great entry point into 3D reconstruction
You don't need expensive hardware to get started — free Colab works fine
For PreserveMy.World, this could be used to get quick depth previews of heritage site photos

GitHub Repo

All my code, research notes, and experiment outputs are in my repo:
https://github.com/HumnaAttique/PMW-day1

Next Steps

I want to try COLMAP next week with real photos, and eventually work toward NeRF or Gaussian Splatting for more realistic heritage site reconstructions.