Overview:

The Mexican Sign Language Alphabet dataset is a comprehensive collection of static signs representing the Mexican Sign Language (LSM) alphabet. This dataset is designed to facilitate research and development in sign language recognition and understanding. It includes 21 distinct signs, covering all letters except J, K, Ñ, Q, X, and Z, which are dynamic signs.

Data Representation:

The dataset comprises images of hand signs, captured against a uniform green background.

The signs are represented as static images, allowing for a clear and standardized view of each sign.

A total of 20 participants were involved in capturing the dataset, ensuring a diverse range of sign representations.

Data Collection:

The data collection process employed a green screen and constant illumination to maintain consistent visual quality across all images.

To capture a wide range of variations, the dataset is divided into three distinct groups: A, B, and C.

Group A contains images with low variation, primarily focusing on light rotations among all degrees of freedom.

Group B features images with rotations along three axes (yaw, roll, pitch), introducing variations in sign orientation.

Group C comprises images with high variation, emphasizing pronounced movements along all degrees of freedom.

Data Partitioning:

The dataset is structured into train and test partitions, ensuring a reliable evaluation of model performance.

The train partition contains data from 18 randomly selected participants, while the test partition includes data from the remaining 2 participants.

Directory Structure:

The root folder contains subdirectories for the three distinct variation groups: lss-abc-A, lsm-abc-B, lsm-C.

Each of these variation groups is further divided into "train" and "test" partitions.

Within the "train" and "test" partitions, you will find subdirectories labeled with class names, representing individual signs (e.g., A, B, C, D, ..., Y).

Inside each class directory, you will find images in jpg format, depicting the corresponding sign. These images are organized for training and evaluation purposes.

Cite by using the DOI https://doi.org/10.5281/zenodo.10067508