SalihHub/blind-assist-tr-image-to-text_and_QA_suitable_UnslothFinetune

该数据集旨在训练多模态模型，以帮助视觉障碍者在室内环境中导航。它包含59,035个图像-文本对，涵盖57个不同的室内场景类别。每个样本包括一张室内场景照片和一个关于场景的土耳其语问答对（例如障碍物检测、地板纹理、物体位置、导航指导）。数据集适用于盲人辅助/室内导航助手、土耳其语图像标注、土耳其语视觉问答以及Unsloth多模态微调（适合`simple_image_text`格式）。数据集分为训练集和测试集，训练集包含56,083个样本，测试集包含2,952个样本。数据集采用CC-BY 4.0许可，允许共享和改编，但需署名。This dataset is designed for training multimodal models to assist visually impaired people in navigating indoor environments. It contains 59,035 image-text pairs covering 57 distinct indoor scene categories. Each sample consists of an indoor scene photograph and a Turkish question-answer pair about the scene (e.g., obstacle detection, floor texture, object location, navigation guidance). The dataset is suitable for blind assist / indoor navigation assistant, image captioning (Turkish), visual question answering (Turkish), and Unsloth multi-modal fine-tuning (suitable for `simple_image_text` format). The dataset is split into train (56,083 samples) and test (2,952 samples) sets. It is licensed under CC-BY 4.0, allowing sharing and adaptation with attribution required.