AI and Digital Workinstructions
Artificial Intelligence (AI) is rapidly conquering our society. But did you know that AI also plays a crucial role in providing digital work instructions? FlandersMake@UHasselt investigates how AI techniques can contribute to efficient and effective instructions for operators.
The AI puzzle piece: Training Data
To train an AI model, we need huge amounts of sample data. Think of images with annotations about the objects in them. Manual annotation is time-consuming and costly. But there is a clever solution: generating images with existing rendering methods, such as those used in game engines. The only problem? We need textured 3D models of all the objects. And the rendered data is often not entirely realistic, something AI models are sensitive to.
Novel View Synthesis: A Creative Solution
This is where novel view synthesis comes in. We create a model based on short videos of an object, which allows us to render that object realistically from any angle. Then we isolate the foreground object from the entire scene and place it in numerous backgrounds. This is how we generate cost-effective, high-quality training data for computer vision models.
Smart Work Instructions
The abundance of new graphics offers tremendous possibilities. During assembly tasks, objects and tools can now be accurately recognized. A digital work instruction system can provide tailored instructions and feedback to operators, increasing efficiency and precision.