This database was created using the 'Sign Language MNIST' database as a reference. It contains 312,000 images of American Sign Language with two additional symbols (space and stop), making it ideal for working on writing or predicting messages. The database consists of 12 mini-databases, each one of them with 1000 samples for each of the 26 signs (24 ASL and 2 auxiliary). Each mini-database has been collected in a different location, using different lightning conditions. The mini-databases are not shuffled, there are firstly 1000 A samples, then 1000 B samples and so on.

Each image has a size of 56x56 pixels squared, and they all are in grayscale, so each pixel contains a value between 0 and 255. The labels have no been set, so each user can set them as he likes. In each mini-database the first 1000 images are A, the next 1000 are B and so on. The order is the following: A, B, C, D, E, F, G, H, I, K, L, M, N, O, P, Q, R, S, T, U, V, W, X, Y, SPACE, STOP.

This database was created in a Bachelor's Thesis. Was used to train a model used for ASL message writing and prediction. Some video-demostrations (different uses of the model, inside the different folders that can be found on the link), the trained model and the database can be found in the following link: https://drive.google.com/drive/folders/1NdFK7Cu9c_eIjNO1OzNUM72DBorH0WJa?usp=sharing