Sign in to view source links and access this dataset
Description
An American Sign Language dataset containing 108,618 videos representing 2,208 distinct words, with each word having a minimum of 30 video examples. The videos were scraped and collected from multiple sources, then preprocessed for consistency and quality. It was uploaded by ZahidYasinMittha and last updated on August 28, 2025.
Use Cases
Train ASL gesture recognition models based on the collection of over 100,000 labeled videos.
Develop video-based machine learning classifiers for sign language words.
Benchmark computer vision models on a large-scale, multi-source video dataset.
Preprocess and standardize video data for consistency in machine learning pipelines.
Strengths
Contains 108,618 videos, providing a substantial volume of data for model training.
Covers 2,208 distinct ASL words, offering a broad vocabulary.
Each word is represented by a minimum of 30 videos, which may help with class balance.
Videos are preprocessed to ensure consistency and are each ≤10 MB, optimized for storage and training.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
The specific sources and collection methodology are not detailed, which may obscure potential biases.
Row count is unknown, which may limit suitability assessment.
Provenance
Source
Multiple scraped sources, as mentioned in the description.
Collection Method
Scraped, collected, and preprocessed from multiple sources.
Freshness
Last updated 2025-08-28 13:02:57; freshness should be verified.
Geography
American
License is unknown; terms of use must be verified before application.