Semantic Attribute Encoding for Industry-Collaborative AI

Description

The transformation of our environments is progressing at a fast pace and save AI systems are thus essential for collaborative systems that operate in human spaces. We propose the “Semantic Encoder”, a CNN model trained on a synthetic images, to address the explainability aspect by extracting semantic descriptions. We can use the extracted information to describe samples or to differentiate between classes. The semantic description can be further used to sort samples or to find a sample with specific properties. We compare the computed semantic features with features extracted by a VGG-16 model and classical image processing methods. We investigate performance by computing ROC and PR curves. We use the semantic parameters to understand what causes good and inaccurate anomaly detection decisions and evaluate the classification quality by examining confusion matrices and classification accuracy.

Period	8 Nov 2024
Event title	European Machine Vision Forum 2024: Challenges and Chances in Computer Vision for Human-Machine Interaction
Event type	Conference
Location	Mulhouse, FranceShow on map
Degree of Recognition	International