Cross-Supervised Object Detection
This addresses the need for more efficient object detection in computer vision by reducing annotation costs, though it is incremental as it builds on existing weakly supervised methods.
The paper tackles the problem of building object detectors for new categories using only image-level annotations, which are cheaper than instance-level ones, by leveraging knowledge from fully labeled base categories. The result is a novel learning paradigm called cross-supervised object detection that improves detection of novel objects in complex scenes like COCO.
After learning a new object category from image-level annotations (with no object bounding boxes), humans are remarkably good at precisely localizing those objects. However, building good object localizers (i.e., detectors) currently requires expensive instance-level annotations. While some work has been done on learning detectors from weakly labeled samples (with only class labels), these detectors do poorly at localization. In this work, we show how to build better object detectors from weakly labeled images of new categories by leveraging knowledge learned from fully labeled base categories. We call this novel learning paradigm cross-supervised object detection. We propose a unified framework that combines a detection head trained from instance-level annotations and a recognition head learned from image-level annotations, together with a spatial correlation module that bridges the gap between detection and recognition. These contributions enable us to better detect novel objects with image-level annotations in complex multi-object scenes such as the COCO dataset.