28 lines
963 B
Markdown
28 lines
963 B
Markdown
# swipe input keyboard for indian languages.
|
|
given a gesture made on the keybard,language chosen-> keyboard layout
|
|
predict the word that matches closest.
|
|
|
|
|
|
## Input
|
|
gesture data -> polygon(shape,size,corners, path), (time, pauses)?, spatial data with word character correlation.
|
|
weighted-vocabulary,corpus for the language,history of gesture-word mappings/corrections for the user.
|
|
language, keyboard layout
|
|
|
|
## Output
|
|
Predict the word
|
|
|
|
## Model
|
|
Structured Output/HMM/CNN?
|
|
|
|
# mnist hand-written digit database -> build application for recognizing full phone numbers(10 digit).
|
|
|
|
## Input
|
|
mnist digit database, generated 10 digit images with random positioning,orientation,scale of individual digit images sampled randomly from the mnist database.
|
|
|
|
## Output
|
|
predict the phone number
|
|
|
|
## Model
|
|
regression model to identify the points where the split for the images has to be made and pass the
|
|
split images to mnist digit recognizer to identify the digit.
|