hagrid-subset
收藏资源简介：
HaGRID手势识别子集是HaGRID（手势识别图像数据集）的一个精选子集，包含24个手势类别，用于训练手势识别模型。数据集总共有19,200张图像，每个类别有800个样本，图像格式为JPEG，平均大小约为302 KB。数据集分为训练集（76%）、验证集（9%）和测试集（15%）。手势类别包括呼叫、不喜欢、拳头、四、抓取、握紧、喜欢、中指、静音、无手势、OK、一、手掌、和平、倒置和平、指向、摇滚、停止、倒置停止、三、三2、三3、两上和倒置两上。数据集结构包括训练、验证和测试目录，以及一个包含19,200条目的注释文件。数据集创建时使用了分层抽样以保持训练/验证/测试比例和类别平衡。数据集验证状态为完全验证，注释覆盖率为100%。许可证为MIT许可证（继承自原始HaGRID数据集）。
The HaGRID Gesture Recognition Subset is a curated subset of the HaGRID (Hand Gesture Recognition Image Dataset), containing 24 gesture categories for training gesture recognition models. The dataset has a total of 19,200 JPEG-format images, with 800 samples per category and an average size of approximately 302 KB. It is split into three subsets: the training set (76%), validation set (9%), and test set (15%). The gesture categories include: call, dislike, fist, four, grab, hold, like, middle finger, mute, no gesture, OK, one, palm, peace, inverted peace, point, rock, stop, inverted stop, three, three2, three3, two fingers up, and inverted two fingers up. The dataset structure includes training, validation, and test directories, as well as an annotation file with 19,200 entries. Stratified sampling was used during dataset creation to maintain the training/validation/test split ratio and ensure category balance. The dataset is fully validated with 100% annotation coverage. The license is the MIT License, inherited from the original HaGRID dataset.
HaGRID Gesture Recognition Subset 数据集概述
数据集基本信息
- 数据集名称：HaGRID Gesture Recognition Subset
- 许可证：MIT License
- 任务类别：图像分类、目标检测
- 标签：手势识别、计算机视觉、手部手势
- 数据规模：10K<n<100K
- 图像总数：19,200
- 手势类别数：24
- 每类样本数：800
- 图像格式：JPEG
- 平均图像大小：约302 KB
数据集划分
| 划分 | 图像数量 | 百分比 |
|---|---|---|
| 训练集 | 14,592 | 76% |
| 验证集 | 1,728 | 9% |
| 测试集 | 2,880 | 15% |
手势类别列表
- call
- dislike
- fist
- four
- grabbing
- grip
- like
- middle_finger
- mute
- no_gesture
- ok
- one
- palm
- peace
- peace_inverted
- point
- rock
- stop
- stop_inverted
- three
- three2
- three3
- two_up
- two_up_inverted
数据集结构
data/ ├── train/ │ ├── call/ (608 images) │ ├── dislike/ (608 images) │ └── ... ├── val/ │ └── ... └── test/ └── ... annotations.csv (19,200 entries)
使用方法
使用 Hugging Face Datasets 加载
python from datasets import load_dataset dataset = load_dataset("YOUR_USERNAME/hagrid-subset") train_data = dataset["train"] val_data = dataset["validation"] test_data = dataset["test"]
使用 Pandas 加载
python import pandas as pd from PIL import Image df = pd.read_csv("hf://datasets/YOUR_USERNAME/hagrid-subset/annotations.csv") train_df = df[df[split] == train]
引用信息
如需使用本数据集，请引用原始 HaGRID 数据集： bibtex @inproceedings{hagrid2022, title={HaGRID--HAnd Gesture Recognition Image Dataset}, author={Kapitanov, Alexander and Kvanchiani, Karina and Nagaev, Alexander and Kraynov, Andrey and Makhliarchuk, Maxim}, booktitle={2022 International Conference on Robotics and Artificial Intelligence (ICRAI)}, year={2022} }
数据集创建与验证
- 创建方式：采用分层抽样方法创建，以保持训练/验证/测试集的比例并确保类别平衡。
- 验证状态：✅ 完全验证，注释覆盖率为100%
