PyLaia
What is PyLaia?
PyLaia is a toolkit for Automatic Text Recognition (ATR) and Keyword Spotting (KWS).
PyLaia is flexible, open-source, device-agnostic, and can be used to express a wide variety of experiments, including (but not limited to) training and inference over Convolutional and Recurrent based deep Neural Network models. The software is extensible and easily configurable and provides a rich set of functional layers with a particular focus on ATR.
History
PyLaia is the successor of Laia. It was developed by 3 members (@jpuigcerver, @mauvilsa, @dmartinalbo) of the Pattern Recognition and Human Language Technology (PRHLT) research center in 2016.
The toolkit was originally developed using Torch. When Torch's development was discontinued in 2017, it became clear that building PyLaia as a second-generation system using PyTorch as its foundation was a logical step. PyLaia was written in 2018 by @jpuigcerver as a Ph.D. thesis experiment and by @carmocca as an undergraduate final project.
Since 2022, three members of TEKLIA (@babadie, @yschneider, @starride) maintain and improve the toolkit.
Get started
Click here to learn more about the original paper.