pzhou10/3D-PC

--- language: - en license: cc-by-4.0 size_categories: - 1K<n<10K task_categories: - image-classification pretty_name: 3D perception challeng dataset_info: - config_name: depth features: - name: image dtype: image - name: split dtype: string - name: label dtype: float64 - name: category dtype: string - name: scene dtype: string - name: setting dtype: int64 - name: img_id dtype: string splits: - name: train num_bytes: 5116762724.6 num_examples: 3848 - name: test num_bytes: 160973952 num_examples: 94 - name: validation num_bytes: 632995269 num_examples: 428 download_size: 6433592072 dataset_size: 5910731945.6 - config_name: vpt-basic features: - name: image dtype: image - name: split dtype: string - name: label dtype: float64 - name: category dtype: string - name: scene dtype: string - name: setting dtype: int64 - name: img_id dtype: string splits: - name: train num_bytes: 9227531453.024 num_examples: 6568 - name: test num_bytes: 160973952.0 num_examples: 94 - name: validation num_bytes: 1055936900.0 num_examples: 730 download_size: 21745516242 dataset_size: 10444442305.024 - config_name: vpt-strategy features: - name: image dtype: image - name: split dtype: float64 - name: label dtype: int64 - name: category dtype: string - name: scene dtype: string - name: setting dtype: int64 - name: img_id dtype: string splits: - name: train num_bytes: 149604226.0 num_examples: 100 download_size: 149610294 dataset_size: 149604226.0 configs: - config_name: depth data_files: - split: train path: depth/train-* - split: test path: depth/test-* - split: validation path: depth/validation-* - config_name: vpt-basic data_files: - split: train path: vpt-basic/train-* - split: test path: vpt-basic/test-* - split: validation path: vpt-basic/validation-* - config_name: vpt-strategy data_files: - split: train path: vpt-strategy/train-* --- ## Project Description - **arXiv:** https://arxiv.org/abs/2406.04138 - **GitHub Repo:** https://github.com/serre-lab/VPT - **Project Page:** https://serre-lab.github.io/VPT ## Summary Visual perspective taking (VPT), the ability to accurately perceive and reason about the perspectives of others, is an essential feature of human intelligence. Deep neural networks (DNNs) may be a good candidate for modeling VPT and its computational demands in light of a growing number of reports indicating that DNNs gain the ability to analyze 3D scenes after training on large static-image datasets. We developed the 3D perception challenge (3D-PC) for comparing 3D perceptual capabilities in humans and DNNs. The 3D-PC is comprised of three 3D-analysis tasks posed within natural scene images: 1. A simple test of object depth order (depth), 2. A basic VPT task (vpt-basic), 3. A version of VPT (vpt-strategy) designed to limit the effectiveness of "shortcut" visual strategies. ## Citation ``` latex @misc{linsley20243dpc, title={The 3D-PC: a benchmark for visual perspective taking in humans and machines}, author={Drew Linsley and Peisen Zhou and Alekh Karkada Ashok and Akash Nagaraj and Gaurav Gaonkar and Francis E Lewis and Zygmunt Pizlo and Thomas Serre}, year={2024}, eprint={2406.04138}, archivePrefix={arXiv}, primaryClass={cs.CV} } ```

原始信息汇总

数据集概述

基本信息

• 语言：英语
• 许可证：CC-BY-4.0
• 大小分类：1K<n<10K
• 任务分类：图像分类
• 美观名称：3D感知挑战

数据集配置

1. 配置名称：depth

   • 特征：
     • image: 图像
     • split: 字符串
     • label: 浮点数
     • category: 字符串
     • scene: 字符串
     • setting: 整数
     • img_id: 字符串
   • 分割：
     • train: 3848个样本，5116762724.6字节
     • test: 94个样本，160973952字节
     • validation: 428个样本，632995269字节
   • 下载大小：6433592072字节
   • 数据集大小：5910731945.6字节

2. 配置名称：vpt-basic

   • 特征：
     • image: 图像
     • split: 字符串
     • label: 浮点数
     • category: 字符串
     • scene: 字符串
     • setting: 整数
     • img_id: 字符串
   • 分割：
     • train: 6568个样本，9227531453.024字节
     • test: 94个样本，160973952字节
     • validation: 730个样本，1055936900.0字节
   • 下载大小：21745516242字节
   • 数据集大小：10444442305.024字节

3. 配置名称：vpt-strategy

   • 特征：
     • image: 图像
     • split: 浮点数
     • label: 整数
     • category: 字符串
     • scene: 字符串
     • setting: 整数
     • img_id: 字符串
   • 分割：
     • train: 100个样本，149604226.0字节
   • 下载大小：149610294字节
   • 数据集大小：149604226.0字节

数据文件

• depth：
  • train: depth/train-*
  • test: depth/test-*
  • validation: depth/validation-*
• vpt-basic：
  • train: vpt-basic/train-*
  • test: vpt-basic/test-*
  • validation: vpt-basic/validation-*
• vpt-strategy：
  • train: vpt-strategy/train-*