Sign language translation is a non-trivial task of interpreting text sequences from video sequences. Like other translation problems, the input word order is usually not the same as the output word order. To solve this, we need a model that takes as input videos containing sign language sentences and outputs a translation text, so this is a sequence to sequence problem
T5-video-transformer.ipynb contains the code for this task.
asl_static_recog.ipynb contains the code for a simple fine-tuned Inception-v3 model
that classifies images containing ASL alphabets
To find out more or to read the paper, check out the SLT project page on my website!