Research challenges

Deep learning has in recent years advanced the field of computer vision and improved performance in e.g. object detection and recognition. Despite these advances the state-of-the-art methods are not yet sufficient to provide reliable results in applications of more complex visual data. Visual Intelligence sets out to tackle the research challenges that limit our user partners to utilize complex visual data in their applications.

Limited training data

Limited training data

Developing new methods to solves complex problems from limited training data.
Context and dependencies

Context and dependencies

Utilizing the best of classical- and machine learning models to unlock the potential in complex image data.
Confidence and uncertainty

Confidence and uncertainty

Developing models that can estimate confidence and quantify uncertainty of their predictions.
Explainability and reliability

Explainability and reliability

Opening the "black box" of deep learning to give explainable and reliable predictions.