Mexican sign language dataset

Mexican sign language, like other sign languages, has its own grammar rules and gestures to denote a word, then the hand gestures for the same word, even in Spanish-spoken countries, can vary. The obtention of videos of MSL signs helps to develop a methodology to translate hand gestures into words. This dataset contains 249 words of the MSL selected from the first words learned in childhood, the words can be grouped into 15 subsets: greetings, time, days of the week, months, school supplies, family, house stuff, adjectives, cuisine, clothes, body parts, vehicles, places, pronouns, verbs, professions and states of México.

From the video clips, we obtain a frame sequence of the hand gesture. The environment in the videos is controlled, we use black fabric for the background and black shirts for the individuals. We use 11 individuals, these aspects help to enhance the regions of interest for the hand gesture formation, the hands, and the face, the movement of one or the two hands, the point of contact with the body, and the face gesture mainly describe the meaning of the sign. By controlling the background and the clothing, we can facilitate the segmentation of the regions of interest and obtain highly discriminating features to classify the words and follow the hand movement.

The dataset includes 249 folders, from the 249 classes selected. For every class, the folder includes subfolders from the individuals. The subfolder contains the frame sequence images of the person making the hand gesture of the corresponding word, in JPG format. The number of images varies depending on the video last.