A tensorflow neural network model to classify to decipher sign language made by hands over six numbers 0 to 5. The trained tensorflow model can be easily downloaded along with its weights. Model has trained over 2160 images over six classes for 2500 iteration over a batch size of 32. It has test_set accuracy of 75.8% which can be improved using regularisation and can further be improved by using convolutional neural network architecture. It can be improved further to facilitate communications from a speech-impaired person to someone who doesn't understand sign language.