Meta-Album SPT: 800 Sports Action Images Across 20 Classes

800 images across 20 sports action classes, resized to 128x128 pixels for few-shot learning. The dataset is a processed subset of the 100-Sports collection, originally gathered from internet searches and deduplicated. It was created for the Meta-Album benchmark by Jilin He in March 2022.

Use Cases

• Benchmarking few-shot image classification models based on the 20 sports categories.
• Training lightweight computer vision models on the 128x128 pixel images.
• Studying domain adaptation for human action recognition using the sports domain.

Strengths

• 800 images provide a defined scale for few-shot learning experiments.
• 20 classes offer a multi-class classification task.
• Images are preprocessed to a consistent 128x128 pixel size.

Limitations

• Column-level documentation is absent; image labels and features must be inferred after download.
• The dataset is a small, curated subset of a larger collection, which may limit generalizability.
• Last update date is unknown; freshness unverified.

Provenance

Source

Processed from the 100 Sports Image Classification dataset by Gerald Piosenka on Kaggle.

Collection Method

Images were gathered from internet searches, deduplicated, and resized for Meta-Album.