Connor Blazek Summary Motivation People with hearing or speech impairment often use sign language to communicate Most people do not understand sign language Results Performed well within the dataset ID: 754713
Download Presentation The PPT/PDF document "American Sign Language Alphabet Recognit..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
American Sign Language Alphabet Recognition
Connor BlazekSlide2
Summary
Motivation
People with hearing or speech impairment often use sign language to communicate
Most people do not understand sign language
Results
Performed well within the dataset
Performed poorly outside of the datasetSlide3
Approach
MobileNet
Pre-trained
Transfer
Learning
Ranks Images based on “confidence”Changed input parameters and datasetsDatasets found on Kaggle3000 images per signAchieved ~ 96% accuracy within datasetSlide4
Approach
Added 2 smaller datasets
Now only 90 images per
sign
Achieved ~ 85% accuracy within dataset
Better with other signs but still very badSlide5
Results
Learning Rate of 0.05 was usually the best
Low Learning Rates caused slower learning an a lower accuracy
High Rates only increase confidenceSlide6
Results
Struggles with:
H
, U
L, T, X
A, E, N (+ M, S)VSlide7
Discussion
Lessons learned
Variety in dataset is important
Learning Rate of about 0.05 yields best accuracy
How to improve
Use more diverse datasetMake sure overfitting has not taken placeSlide8
References
Mekala
, Priyanka & Gao, Ying & Fan, Jeffrey &
Davari
, A. (2011). Real-time sign language
recognition based on neural network architecture. Proceedings of the Annual Southeastern Symposium on System Theory. 195 - 199. 10.1109/SSST.2011.5753805. Zee, S. (2018, September 20). Whose Sign Is It Anyway? AI Translates Sign Language Into Text. Retrieved October 7, 2018, from https://blogs.nvidia.com/blog/2017/05/11/ai-translates-sign-language/
Datasets:https://www.kaggle.com/grassknoted/asl-alphabethttps://
www.kaggle.com/danrasband/asl-alphabet-testhttps://www.kaggle.com/augustohaja/asl-lucas