1 paper
Chi Ruan, Jiying Zhao, Wenhu Chen
Although open-vocabulary object detectors can generalize to unseen categories, they still rely on predefined textual prompts or classifier heads during inference. Recent generative…