Stanford 40 Actions: Human Action Images for Computer Vision

The Meta-Album Stanford 40 Actions Dataset (Extended) contains 3,389 images of humans performing 40 distinct actions, preprocessed to a uniform 128x128 pixel size. It is derived from the original Stanford 40 Actions dataset, which was created by researchers including Bangpeng Yao and Li Fei-Fei. The Meta-Album version was created by Jilin He and released in March 2022.

Use Cases

• Benchmarking few-shot image classification models based on the 40 distinct action classes.
• Training models for human action recognition in still images.
• Evaluating model performance on a standardized, preprocessed image dataset with uniform 128x128 resolution.

Strengths

• Contains 3,389 images across 40 distinct action classes, providing a multi-class classification task.
• Images are preprocessed to a uniform 128x128 pixel size, ensuring consistency for model input.
• Clear provenance linking to the original Stanford 40 Actions dataset and associated research paper.

Limitations

• The dataset is a subset of the original 9,532 images; the selection criteria for the 3,389 images are not detailed.
• Column-level documentation and sample data are unavailable, making detailed feature inspection impossible before download.
• The last update date for this specific version on the platform is unknown.

Provenance

Source

Stanford 40 Actions dataset (http://vision.stanford.edu/Datasets/40actions.html), preprocessed and released by Meta-Album.

Collection Method

Images were resized to 128x128 pixels using OpenCV as part of the Meta-Album preprocessing pipeline.