Open-Vocabulary Detection

Open-vocabulary detection moves object detection beyond closed benchmark categories toward more flexible real-world scenarios. The model can attempt to recognize object types described in natural language or not explicitly trained beforehand. This is highly valuable for dynamic catalogs, robotics, and broad media indexing. It clearly demonstrates the direct impact of vision-language modeling on perception systems.