HuggingFaceM4/clevr

该数据集是Visual Question Answering (VQA)数据集，包含了大量的问题、答案和图像。具体统计信息如下：训练集包含443,757个问题和4,437,570个答案，验证集包含214,354个问题和2,143,540个答案，测试集包含447,793个问题。图像数量方面，训练集有82,783张图像，验证集有40,504张图像，测试集有81,434张图像。此外，未明确提及的testdev子集包含107,394个问题和36,807张图像。

This dataset is a Visual Question Answering (VQA) dataset that encompasses a vast collection of questions, answers, and accompanying images. Specific statistical details are outlined below:
The training split comprises 443,757 questions and 4,437,570 answers, paired with 82,783 unique images.
The validation split includes 214,354 questions, 2,143,540 answers, and 40,504 corresponding images.
The test split contains 447,793 questions and 81,434 images.
Furthermore, the testdev subset, which is not explicitly mentioned in the preceding context, consists of 107,394 questions and 36,807 images.

原始信息汇总

数据集概述

训练集（Train）

• 问题数量: 443,757
• 答案数量: 4,437,570
• 图像数量: 82,783

验证集（Val）

• 问题数量: 214,354
• 答案数量: 2,143,540
• 图像数量: 40,504

测试集（Test）

• 问题数量: 447,793
• 图像数量: 81,434

测试开发集（Testdev）

• 问题数量: 107,394
• 图像数量: 36,807