Skip to main content

Our technology

Accelerated AI Model Builder is a self-contained, powerful tool offering ground-truth data generation and autonomous detection model building as a (web) service. It is applicable on any image or video data and in any domain.

Users can generate detectors working on image data, while the detection target can be freely defined. By using innovative techniques, Accelerated AI Model Builder proposes a minimized number of images to be annotated by experts and builds an optimized & tested detection model using received annotations.

As a result, Accelerated AI Model Builder provides in a very efficient & fast way detectors with significantly reduced human input, avoiding that users need to review a high number of images or video sequences.

Project members

APA style publications

  • Kushnir, Dan; Venturi, Luca. "Diffusion Based Deep Active Learning," arXiv, (23 March 2020). (Link)

  • Kushnir, Dan; Mirabelli, Benjamin. "Active Community Detection with Maximal Expected Model Change," Proceedings of Machine Learning Research, Volume 108 (26-28 August 2020). (Link)

  • Kushnir, Dan. "Active-transductive Learning With Label-adapted Kernels," KDD '14: Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining (August 2014).(Link)