TECHSALO
CASE STUDY / AR / VR / IMMERSIVE COMMERCE

AR COMMERCE
BEFORE AR BECAME A KIT.

A framework-independent computer-vision and 3D platform that delivered true-scale product visualisation across the web, iOS and Android.

PLATFORMS
Web · iOS · Android
ERA
2016–2018
ENGINE
Custom CV + 3D
Explore the system ↓
THE ENGINEERING MANDATE

MAKE DIGITAL PRODUCTS FEEL PHYSICALLY PRESENT.

In 2016, mobile operating systems did not provide a dependable cross-platform AR layer. Retailers still needed shoppers to place true-scale products in a live camera view.

TechSalo engineered the missing platform: camera access, visual tracking, pose estimation, real-time rendering and a production 3D-asset pipeline.

3runtime platforms
0packaged AR SDK dependencies
1shared 3D catalogue
01 / CUSTOM AR RUNTIME

FROM CAMERA FRAME TO STABLE PLACEMENT.

The runtime combined platform camera feeds, a shared C++ vision core and platform-specific renderers. Every stage was designed to recover gracefully as device capability, light and tracking confidence changed.

01Capture

Read camera frames and device motion through browser or native APIs.

02Detect

Find a known marker or planar target using keypoints and descriptors.

03Estimate pose

Calculate position and orientation with homography and PnP geometry.

04Render

Anchor a true-scale model, apply lighting and smooth each movement.

05Measure

Capture product and session events without retaining camera imagery.

02 / ONE EXPERIENCE, THREE RUNTIMES

SHARED LOGIC. NATIVE PERFORMANCE.

BROWSER RUNTIMEWEB

Camera access through getUserMedia, tracking separated from the interface, and WebGL/Three.js rendering with capability-based fallbacks.

JAVASCRIPT · WEB WORKERS · THREE.JS · WEBGL
NATIVE RUNTIMEiOS

AVFoundation supplied frames to the shared vision core; OpenGL ES and Metal rendered models independently of ARKit.

C++ · OPENCV · AVFOUNDATION · METAL
NATIVE RUNTIMEANDROID

Camera2 and device sensors connected through JNI to shared tracking logic, with OpenGL ES handling the scene.

C++ · OPENCV · CAMERA2 · OPENGL ES
03 / CLOUD 3D ASSET PIPELINE

HEAVY SOURCE MODELS. LIGHTWEIGHT EXPERIENCES.

01Reference

Collect CAD, dimensions and product photography.

02Model

Clean or rebuild geometry in correct real-world units.

03Surface

Create UVs and bake material and normal detail.

04Optimise

Retopologise, create LODs and compress delivery files.

05Validate

Check scale, load time and tracking on target devices.

04 / PLATFORM CAPABILITIES

THE FULL STACK BEHIND ONE NATURAL INTERACTION.

  • Camera calibration and frame preprocessing
  • Marker and planar-image feature detection
  • Homography and 6DoF pose estimation
  • Sensor-assisted smoothing and tracking recovery
  • True-scale placement, rotation and product variants
  • Cloud conversion, LOD generation and CDN delivery
  • Multi-tenant catalogue and engagement analytics
05 / ENGINEERING OUTCOMES

REUSABLE AR IP. NO PLATFORM LOCK-IN.

Independent

Reusable AR intellectual property without dependence on a vendor framework.

Consistent

Product scale and placement aligned across supported devices and runtimes.

Resilient

Fallbacks handled lower tracking confidence and constrained device capability.

Centralised

One publishing workflow served multiple retail catalogues and applications.

TECHSALO / EXPERIENCE ENGINEERING

BUILD THE INTERACTION
YOUR PLATFORM NEEDS.

Computer vision, 3D, web and mobile engineering—designed as one production system.

Start a conversation ↗