Document AI 包括了许多数据科学的任务,包括图像分类、图像转文本 (image to text)、文档回答 (document question answering) 、表格回答 (table question answering) 以及视觉回答 (visual question answering) 。 本文将向大家介绍 Document AI 中的六种不同使用场景,以及它们的最佳开源模型,之后重点分三个方面介绍了...
Visual question answering answer a question about the image, given an image and a question Multimodal pipeline(task=“vqa”) Document question answering answer a question about a document, given an image and a question Multimodal pipeline(task=“document-question-answering”) Image captioning generate...
Visual question answering answer a question about the image, given an image and a question Multimodal pipeline(task=“vqa”) Document question answering answer a question about a document, given an image and a question Multimodal pipeline(task=“document-question-answering”) Image captioning generate...
Document Question Answering document-question-answering Visual Question Answering visual-question-answering Image-to-Text image-to-text 大语言对话模型 您可以在官方库tasks中选择要部署的大语言对话模型,并获取大语言对话模型的MODEL_ID(模型ID)、TASK(模型对应的TASK)、REVISION(模型版本)的值并保存到本地。目前...
"document-question-answering": will return a DocumentQuestionAnsweringPipeline. "feature-extraction": will return a FeatureExtractionPipeline. "fill-mask": will return a FillMaskPipeline:. "image-classification": will return a ImageClassificationPipeline. "image-segmentation": will return a ImageSegmenta...
Multimodal(多模态):Feature Extraction(特征提取)、Text-to-Image(文本到图像)、Visual Question Answering(视觉问答)、Image2Text(图像到文本)、Document Question Answering(文档问答) Tabular(表格):Tabular Classification(表分类)、Tabular Regression(表回归) ...
document-question-answering Feature Extraction feature-extraction Image Feature Extraction image-feature-extraction Image-to-Text image-to-text Mask Generation mask-generation Visual Question Answering visual-question-answering 示例 下面是一个用户层面创建步骤 进入到huggingface模型详情页,查看对应的模型ID/模型TAS...
AutoModelForDocumentQuestionAnswering AutoModelForImageClassification AutoModelForImageSegmentation AutoModelForInstanceSegmentation AutoModelForMaskedImageModeling AutoModelForMaskedLM AutoModelForMultipleChoice AutoModelForNextSentencePrediction AutoModelForObjectDetection ...
History 15 Commits analysis clean_and_create create_only_with_pdfs florence_2_dataset generation zero_shot_exp .gitignore LICENSE README.md docmatix_thumbnail.png README MIT license Docmatix Docmatix is a comprehensive dataset designed for Document Visual Question Answering (DocVQA). It provides...
对一个document 或一篇文章生成摘要。 一个summarization 任务的数据集是 CNN / Daily Mail dataset, 包含了很多新闻文章,就是为了summarization任务而打造。如果你想要基于summarization任务微调,不同的方法在下面的文档中描述了。 下面是使用模型和tokenizer进行summarization的流程: 根据checkpoint name 初始化 model 和 ...