SGN and EQA

该数据集专注于语言条件下的视觉导航任务，特别是语义目标导航（SGN）和具身问答（EQA）。该数据集旨在评估模型在SGN和EQA任务上的表现，结果表明，双重注意力模型显著优于基线模型。数据集规模包括简单设置下的1000万帧和困难设置下的5000万帧，任务旨在联合学习语义目标导航和具身问答。

This dataset focuses on vision-and-language navigation tasks, specifically Semantic Goal Navigation (SGN) and Embodied Question Answering (EQA). It is designed to evaluate model performance on these two tasks, and the experimental results demonstrate that the dual-attention model significantly outperforms baseline models. The dataset includes 10 million frames under the simple setup and 50 million frames under the challenging setup, with its tasks aiming to jointly learn Semantic Goal Navigation and Embodied Question Answering.