BanglaLekhaImageCaptions
收藏arXiv2025-09-30 收录
下载链接：
https://data.mendeley.com/datasets/rxxch9vw59/2
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为BanglaLekhaImageCaptions，包含了9,154张来自公共领域的图片，这些图片与孟加拉文化相关，适用于训练和验证图像字幕生成模型。需要注意的是，该数据集中包含人类偏见，可能会影响模型的性能，因此在进行模型训练前，需要进行预处理以降低噪声。该数据集的规模为9,154张图片，所针对的任务是图像字幕生成。
The dataset is named BanglaLekhaImageCaptions. It includes 9,154 public-domain images related to Bengali culture, and is designed for training and validating image captioning models. It is worth noting that this dataset contains human biases, which may adversely affect model performance. Therefore, preprocessing is required to reduce noise before model training. This dataset has a total of 9,154 images, and its targeted task is image captioning.
提供机构：
Public Domain
