YOLOOC: YOLO-based Open-Class Incremental Object Detection with Novel Class Discovery
This addresses the challenge of incremental learning in object detection for real-world applications where novel classes appear without prior supervision, though it is incremental as it builds on YOLO architecture.
The paper tackles the problem of open-world object detection where models must detect novel classes at inference and learn them incrementally without forgetting known classes, proposing YOLOOC which uses label smoothing to prevent overconfidence and discovers novel classes effectively in a new benchmark.
Because of its use in practice, open-world object detection (OWOD) has gotten a lot of attention recently. The challenge is how can a model detect novel classes and then incrementally learn them without forgetting previously known classes. Previous approaches hinge on strongly-supervised or weakly-supervised novel-class data for novel-class detection, which may not apply to real applications. We construct a new benchmark that novel classes are only encountered at the inference stage. And we propose a new OWOD detector YOLOOC, based on the YOLO architecture yet for the Open-Class setup. We introduce label smoothing to prevent the detector from over-confidently mapping novel classes to known classes and to discover novel classes. Extensive experiments conducted on our more realistic setup demonstrate the effectiveness of our method for discovering novel classes in our new benchmark.