dddraxxx/sealvqa-vaw-3000

该数据集包含了图像和与图像相关的多种信息，如数据来源、奖励模型、额外信息等。奖励模型中包含了地面真实信息，而额外信息则包括了答案、边界框格式、对话轮数、数据划分、数据集类型、是否具有目标、图像路径、索引、原始对话、问题、问题ID、搜索次数和目标实例等。数据集分为训练集，配置文件提供了训练集的数据路径。

The dataset includes images and various related information such as data source, reward model, extra information, etc. The reward model contains ground truth information, while the extra information includes answer, bounding box format, number of conversation turns, data split, dataset type, whether it has targets, image path, index, original conversations, question, question ID, number of searches, and target instances. The dataset is split into a training set, and the configuration file provides the data path for the training set.