The meat/most of the code can be found in the asl_recognition.ipynb notebook. The data directory holds the original data but preprocessed for the asl_recognition.ipynb notebook. However, the original data can be found at www.cvssp.org/FingerSpellingKinect2011/fingerspelling5.tar.bz2 or alternatively can be downloaded by running the data_preprocessing.ipynb notebook which will first download the images and then proceed to organize and preprocess the data for later.
The capstone's final paper (American Sign Language Handshape Detection.pdf) can be found in the final_paper directory. The proposal can be found in the proposal directory. A zip file final_submission.zip has been created for convenience containing both papers.
Other code and resource has been provided in this repo particularly the evaluation directory which holds some of the helping code used to finalize some analysis for the final paper. The original benchmark paper can be found at the researcher's website http://empslocal.ex.ac.uk/people/staff/np331/index.php?section=FingerSpellingDataset.
Libraries used include pandas, matplotlib, numpy, PIL, seaborn, sklearn, keras, and tensorflow.