NoCaps

Dubbed NoCaps, for novel object captioning at scale, NoCaps consists of 166,100 human-generated captions describing 15,100 images from the Open Images validation and test sets. The associated training data consists of COCO image-caption pairs, plus Open Images image-level labels and object bounding boxes. Since Open Im

Papers using NoCaps (4)

Are Neuro-Inspired Multi-Modal Vision-Language Models Resilient to Membership Inference Privacy Leakage?2025The Telephone Game: Evaluating Semantic Drift in Unified Models2025A Frustratingly Simple Approach for End-to-End Image Captioning2022 · 12 citesRAVEN: Multitask Retrieval Augmented Vision-Language Learning2024 · 1 cites