Automated Classification and Feature Extraction for Experimental Materials Data
Overview
This service provides advanced automated classification and feature extraction systems for experimental data on material properties. Going beyond traditional statistical methods or standard image analysis techniques, the service utilizes deep neural networks to identify and categorize features with higher speed and accuracy. The service mostly focuses on image-like data (e.g., microscopy images, sensory images), which are often high-dimensional and unstructured. Deep learning models can automatically learn hierarchical representations from raw data, extracting meaningful patterns without manual feature engineering. The service bridges the gap between research of complex features in data and industrial application. It allows companies to automate material identification, especially if they contain defects/impurities that were previously found manually or relied on slower image analysis techniques. This leads to faster throughput in production lines and enhanced scalability for large-scale industrial tasks, with a specific focus on identifying material impurities and anomalies. The service is based on fully supervised deep-learning computer-vision models, especially a modified version of the original U-Net algorithm that can handle multi-channel input (of shape: height, width, channel), trained on labeled examples in a supervised manner. These models learn detailed visual representations of material appearance, enabling accurate detection and classification of defects such as surface damage, shape irregularities, texture anomalies, or structural deviations.
Outcome
Depending on the customer's target goals, outputs may include classification scores, predicted boundaries around impurities or defects, and per-pixel probability maps for material characterization.
Service Card
| Category | AI products |
| Domain | Industry, Research and Development, Smart Manufacturing |
| Vertical | Green Transition |
| Target users | R&D in High-Tech Companies, Academic Research, Public Institutions |
| Service type | Model/Algorithms |
| Outputs | Fine-tuned model for training; per-pixel probability score for particular task plus feature classification score for inference. |
| Requirements | Requires a workstation with an NVIDIA GPU node (with at least 8GB RAM, RTX-30 or higher). For training: sufficient number of annotated training images prepared for segmentation task (i.e., per-image and per-pixel annotation) and prepared within the TensorFlow framework. |
| Limitations | |
| Availability | Q1/2027 |
| Licensing | MIT and CC BY-NC-SA 4.0 |
| Visual content | assets/service-visuals/ung-materials.png |
| Visual caption | An example of AFM scan of P3HT (Poly-3-hexylthiophene), semiconducting conjugated polymer used in organic electronics. Deep learning algorithm can extract information on layer structure, morphology and the orientation of grains/islands |