Unlike traditional anomaly detection systems, which often depend on cloud infrastructure or complex hardware setups, GoPxL Anomaly Detector allows for on-site deployment and training. Users can train and deploy 3D defect detection models directly on the GoMax Smart Vision Accelerator, eliminating the need for cloud-based storage or external processing power. This leads to faster model updates, reduced latency, and improved overall efficiency in production.
From Part to Surface Inspection
One of the key strengths is its versatility across various use cases. The tool can be applied to a wide range of inspection tasks, including:
- Part Inspection: For manufacturers working with complex machined or cast parts, GoPxL Anomaly Detector can quickly detect defects and irregularities. Traditional methods require complex thresholds, CAD files, and cloud-based resources. With the new sofwtare, users can train AI models directly on the production line using a simple dataset of OK and NG (non-good) parts. No cloud connection or CAD data is needed, just the parts themselves.
- Surface Inspection: The new software also excels in surface inspection tasks. For industries like flooring and building materials, it combines traditional and AI-based tools to detect defects on large surfaces. By using surface edge and feature tools to mask certain areas, the tool can focus on specific regions to detect defects. Anomaly detection using 2D intensity data ensures that surface defects, whether on boards or floors, are quickly identified.
Key Features and Benefits
- Train on 2D Intensity or 3D Height Map Data: Whether you’re inspecting simple or complex parts, GoPxL Anomaly Detector allows users to choose between 2D intensity data or 3D surface data for training models. A 3D visualizer improves labelling accuracy during the training process, ensuring precise results.
- Preview Functionality Without a License: Users can evaluate the anomaly detection capabilities in replay mode without a license. This lets you explore the software’s full potential before making any financial commitments.
- Synthetic and Augmented Data Generation: One of the standout features of Anomaly Detector is its ability to generate synthetic and augmented data, reducing the need for a large dataset. This feature improves model performance by providing more data for training without requiring a significant number of images.
- Integrated AI Modelling Workflow: Anomaly Detector’s AI workflow is built directly into the software, allowing users to train, test, and deploy models without moving files between different applications. This integrated approach saves time and ensures a smoother process from model creation to deployment.
- On-Site Training and Inference: With Anomaly Detector, data is trained directly on the production line, eliminating the need to transfer sensitive data to the cloud or other systems. This not only reduces data security risks but also ensures faster training and inference.
- Assisted Labeling for Fast Iterations: After training a model, Anomaly Detector’s assisted labeling tool helps speed up the labeling process for new production data. This makes it easy to iterate on the model and adapt it to real-world conditions quickly.
- Advanced Customization with Python GDK: For advanced users, Anomaly Detector offers a Python-based GDK for creating custom workflows, integrating open-source tools, and pulling measurement thresholds from external files. This level of flexibility is ideal for teams with specific needs.
License-free Evaluation Mode
GoPxL Anomaly Detector integrates with a range of Gocator smart sensors for scan data acquisition, including 3D Line Profiler, Snapshot, and Line Confocal when powered by the GoMax. The Anomaly Detector software can be downloaded directly from the LMI website and is available free of charge, with a licensed LMI Dongle required to run it with LIVE sensors. However, for those looking to evaluate the software, a license-free evaluation mode is available for testing in Replay Mode. This gives users the opportunity to experience the tool’s full capabilities without committing to a license right away.
