FG-OVD

该数据集是一个评估开放词汇目标检测器的能力的基准套件，特别关注于理解物体及其部件的细粒度属性，这些属性的难度各不相同。数据集还包括一个动态词汇生成协议，并在多种条件下评估了多个最先进的目标检测模型，其中包括具有挑战性的负样本设置。该任务的目的是开放词汇目标检测。

This dataset is a benchmark suite for evaluating the capabilities of open-vocabulary object detectors, with a particular focus on understanding fine-grained attributes of objects and their components, which vary in difficulty. The dataset also includes a dynamic vocabulary generation protocol, and evaluates multiple state-of-the-art object detection models under diverse conditions, including challenging negative sample settings. The core task targeted by this benchmark is open-vocabulary object detection.

搜集汇总

背景与挑战

背景概述

FG-OVD是一个专注于细粒度开放词汇对象检测的数据集，包含训练集、验证集和一个改进的NoctOWL模型，用于评估和提升细粒度对象检测性能。数据集遵循LVIS格式，并提供了详细的安装和使用说明。