In code/ you will find all python files used to train and evaluate
our models. Code for the neural networks is also included in lstm.py, rnn.py,
and rownet.py.

data/ contains a sample of the data for a screwdriver. The entire data set is
around 10GB not including the video files used to create the images.
All versions of our collected images types are present: color, depth, raw image, and point cloud.

speech.txt and text.txt are both small tsv files of sample description
collections. The speech wav_id field refers to the .wav files in the same directory.
the selected text and speech transcriptions were collected from the same
view of the screwdriver. It is clear from the last speech transcription
that the transcription method isn't perfect, while the text samples are
all consistent and clear.