VEM.AI: Vision-enabled Manufacturing - Accelerating vision model development for Defect Detection and Automated Material Handling through AI-Augmented Datasets.

Purpose and goal

VEM.AI is an R&D initiative using generative AI to create synthetic data for vision systems in automotive manufacturing. This enables rapid prototyping for two key areas: defect detection for quality assurance and automated material handling to boost efficiency. The project´s goal is to implement and validate generative AI models that produce training datasets for industrial machine vision applications.

Expected effects and result

The project aims to achieve the following goals: * Development of a user-friendly digital platform to empower vision engineers to independently generate high-quality synthetic training data * Optimization of Repli5´s generative models, focus on reducing fine-tuning and inference costs. * The project will demonstrate the practical application of Repli5´s technology in real-world automotive settings, showcasing its ability to deliver cost-effective and robust AI solutions.

Planned approach and implementation

The project is run through 7 structured work packages (WP1: Requirements & Data Procurement, WP2: Generative RGB Model Development, WP3: Multimodal Generative Model Development, WP4: Stereo & Multicamera Augmentations, WP5: Training, Evaluation & Validation, WP6: Digital Platform Development och WP7: Project Management & Business Development) with clear responsibilities for each WP. Collaboration with industry partners ensures knowledge transfer and the practical usage of the project´s results