Tuteldove/coco_vqa_small_dataset
收藏资源简介:
--- viewer: false --- # ## Overview - `Vision question Answer (VQA) dataset`: VQA is a new dataset containing open-ended questions about images. These questions require an understanding of vision, language and commonsense knowledge to answer. - `Reference`: Split into small-train & small-val dataset from https://huggingface.co/datasets/Graphcore/vqa validation dataset ## Dataset Structure ### Data Instances A data point comprises an image and its object annotations. ``` {'question': 'Where is he looking?', 'question_type': 'none of the above', 'question_id': 262148000, 'image_id': 'images/COCO_val2014_000000262148.jpg', 'answer_type': 'other', 'label': {'ids': ['at table', 'down', 'skateboard', 'table'], 'weights': [0.30000001192092896, 1.0, 0.30000001192092896, 0.30000001192092896]}} ``` ### Data Fields - `question`: the question to be answered from the image - `question_type`: - `image_id`: the path to the image the question refers to - `answer_type`: - `label`: the annotations - `ids`: - `weights`: ### Data Splits - `Training dataset` (1169) - `Val dataset` (100) ## Usage ``` from datasets import load_dataset dataset = load_dataset("SIS-2024-spring/coco_vqa_small_dataset") ```
--- viewer: false --- # ## 概览 - **视觉问答(Vision Question Answer, VQA)数据集**:VQA是一类面向图像的开放式问答数据集,此类问题的解答需要结合视觉认知、语言理解与常识知识。 - **参考来源**:本数据集的小训练集与小验证集拆分自https://huggingface.co/datasets/Graphcore/vqa 验证集。 ## 数据集结构 ### 数据实例 单个数据样本包含一幅图像及其目标标注信息。 {'question': 'Where is he looking?', 'question_type': 'none of the above', 'question_id': 262148000, 'image_id': 'images/COCO_val2014_000000262148.jpg', 'answer_type': 'other', 'label': {'ids': ['at table', 'down', 'skateboard', 'table'], 'weights': [0.30000001192092896, 1.0, 0.30000001192092896, 0.30000001192092896]}} ### 数据字段 - `question`:用于基于图像进行解答的问题 - `question_type`: - `image_id`:问题所指向的图像路径 - `answer_type`: - `label`:标注信息 - `ids`:候选答案标识列表 - `weights`:各候选答案的权重值 ### 数据划分 - `训练集`(共1169条样本) - `验证集`(共100条样本) ## 使用方法 from datasets import load_dataset dataset = load_dataset("SIS-2024-spring/coco_vqa_small_dataset")
数据集概述
数据集名称
- Vision question Answer (VQA) dataset:这是一个包含关于图像的开放式问题的数据集,需要视觉、语言和常识知识来回答。
数据集来源
- 该数据集是从Graphcore/vqa验证数据集中划分出的小型训练集和小型验证集。
数据集结构
数据实例
- 每个数据点包括一张图片及其对象标注。
json { "question": "Where is he looking?", "question_type": "none of the above", "question_id": 262148000, "image_id": "images/COCO_val2014_000000262148.jpg", "answer_type": "other", "label": { "ids": ["at table", "down", "skateboard", "table"], "weights": [0.30000001192092896, 1.0, 0.30000001192092896, 0.30000001192092896] } }
数据字段
- question:图像相关的问题。
- question_type:问题类型。
- image_id:图像的路径。
- answer_type:答案类型。
- label:标注信息
- ids:标注的类别。
- weights:类别权重。
数据划分
- 训练集:包含1169个样本。
- 验证集:包含100个样本。




