Contact person: Pocket 3D Scanner App

Phone:Show

Email:Show

Location: Patiāla, India

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"I am building a cross-platform mobile application that lets anyone capture a real-world object, room, or scene with nothing more than their phone camera and instantly turn that capture into an accurate, scaled 3D model ready for AR/VR use. The core challenge is to reach high-precision results without relying on LiDAR; instead we will lean on advanced computer-vision pipelines, neural-network depth estimation, and clever photogrammetry post-processing.

SScope of Work

• Develop the capture workflow for both iOS (ARKit + AVCaptureMultiCamSession) and Android (Camera2 / CameraX with LOGICAL_MULTI_CAMERA), ensuring synchronized dual-camera frame acquisition with guidance for optimal coverage and lighting.
• Implement or integrate a stereo depth-estimation pipeline, beginning with calibrated dual-camera intrinsics/extrinsics to generate high-quality disparity maps, refined into per-pixel depth.
• Feed the disparity-derived depth maps into a 3D reconstruction pipeline capable of producing watertight meshes with photorealistic texture mapping.
• Guarantee metrically accurate reconstructions (scale derived from the fixed stereo baseline) suitable for AR/VR by exporting models in glTF / USDZ for direct integration into Unity, Unreal, or WebXR.
• Provide an on-device preview so the user can verify scale fidelity, mesh quality, and texture consistency before saving or sharing.
• Offer a lightweight cloud fallback for computationally heavy reconstructions (e.g., Poisson meshing, multi-view texturing), ensuring seamless asset hand-off and return.

Acceptance Criteria

A calibrated dual-camera scan of a 2 m × 2 m test cube must report each edge within 1 % of its actual length, leveraging stereo baseline for true scale recovery.

End-to-end processing — from first frame capture to shareable AR-ready file — must complete under three minutes on flagship phones (A-series Bionic / Snapdragon 8+ class).

The same codebase must compile and run with feature parity on iOS 16+ and Android 11+, with stereo capture working on all supported dual-camera devices.

Deliverables include source code, build instructions, and a brief technical report on the stereo depth-estimation and reconstruction approach." (client-provided description)