March 4, 2025
Speaker: Eric Martinson
Open vocabulary object detection is the problem of detecting objects in images best associated with an arbitrary search string. Algorithms like Contrastive Language Image Pre-training (CLIP) and Yolo-World are a radical paradigm shift away from classic machine learning training methods with fixed object categories and hand-labeled datasets. This presentation examines current work in the domain at LTU, recognizing changes in the environment and finding targets with a mobile robot, while also proposing future areas of research and examining their impact on industry.