axxkaya/UVT-Terminological-based-Vision-Tasks
收藏Hugging Face2025-02-25 更新2025-02-15 收录
下载链接：
https://hf-mirror.com/datasets/axxkaya/UVT-Terminological-based-Vision-Tasks
下载链接
链接失效反馈官方服务：
资源简介：
UVT Explanatory Vision Tasks数据集包含12百万个图像输入到解释性指令到输出的三元组，用于训练能够理解和执行视觉任务的模型，特别是在零样本任务泛化方面。数据集中的每个任务都通过详细的自然语言描述来定义，而不是传统的术语定义，从而帮助模型在没有见过的新任务上实现泛化。
The UVT Explanatory Vision Tasks dataset consists of 12 million image input → explanatory instruction → output triplets designed for training models to understand and perform vision tasks, especially in terms of zero-shot task generalization. Each task in the dataset is defined through detailed natural language descriptions rather than traditional terminological definitions, aiding the model in generalizing to unseen tasks.
提供机构：
axxkaya
