AI 3D Model Developer for Jewel

**Upwork Job Post: AI Engineer / Computer Vision Expert for 2D-to-3D South Indian Jewelry Generation** ### Project Overview We are building an AI-powered virtual jewelry try-on platform focused on South Indian jewelry. We are looking for an experienced AI/ML Engineer, Computer Vision Engineer, or 3D AI Specialist to develop a system that can automatically generate high-quality 3D jewelry models from 2D images. The project includes dataset creation, synthetic data generation, model training, and deployment of a production-ready AI pipeline. ### Project Goals The selected freelancer will: * Build a dataset of South Indian jewelry images * Create synthetic training data to improve model accuracy * Develop and train AI models capable of converting 2D jewelry images into 3D representations * Generate photorealistic, high-detail 3D jewelry assets * Optimize the solution for scalability and future production deployment ### Scope of Work #### Phase 1: Dataset Creation * Collect and curate South Indian jewelry datasets * Categorize jewelry types: * Necklaces * Haram * Chokers * Earrings * Jhumkas * Bangles * Vanki * Maang Tikka * Temple Jewelry * Perform image cleaning and labeling * Create metadata structure #### Phase 2: Synthetic Data Generation * Generate high-quality synthetic jewelry images * Create multiple views: * Front * Side * 45-degree angle * Top * Close-up details * Generate variations: * Gold * Antique Gold * Diamond * Ruby * Emerald * Pearl #### Phase 3: AI Model Development Develop one or more of the following: * Multi-view reconstruction models * Neural Radiance Fields (NeRF) * Gaussian Splatting * Diffusion-based 3D generation * Point Cloud Generation * Mesh Reconstruction Expected output: * OBJ * GLB * FBX * PLY formats. #### Phase 4: Model Training & Optimization * Train custom models * Fine-tune open-source models where appropriate * Improve: * Geometry accuracy * Stone placement accuracy * Texture quality * Gold reflection realism #### Phase 5: Deployment * Create inference pipeline * API-based architecture preferred * Cloud deployment experience is a plus * GPU optimization ### Required Skills Must Have: * Python * PyTorch * Computer Vision * Deep Learning * 3D Reconstruction * Mesh Processing * OpenCV * CUDA Preferred Experience: * NeRF * Gaussian Splatting * Instant-NGP * Stable Diffusion * TripoSR * Wonder3D * Zero123 * Meshy AI * Blender Automation * AWS or GCP ### Deliverables * Complete training dataset * Synthetic data generation pipeline * Trained AI model * Source code * Model documentation * Deployment guide * API endpoints for inference * Sample generated 3D jewelry models ### To Apply Please provide: 1. Similar AI/3D reconstruction projects completed 2. Portfolio or GitHub links 3. Technologies you recommend for this project 4. Estimated timeline 5. Estimated cost 6. Experience with NeRF, Gaussian Splatting, or 2D-to-3D generation ### Success Criteria * Generate accurate 3D South Indian jewelry from a single or multiple 2D images * Preserve intricate design details * Produce production-ready 3D assets suitable for virtual try-on applications * Achieve consistent results across multiple jewelry categories **Project Type:** Fixed Price (Preferred) **Experience Level:** Expert **Duration: 1 month **Budget:** Open to proposals based on expertise and deliverables.