dynamic gesture about sign-language HOT 9 OPEN

I think what he meant was, to recognize dynamic gestures such as J and Z... You need to go for 3d conv nets and train using video datasets for that.

from sign-language.

I think the application is recognizing gestures dynamically (gestures can be changed in the video), though that gesture should be stored in the database and also trained on the neural net.

from sign-language.

I think 3d conv nets can recognize only static gesture or image right? for dynamic need to use lstm or rnn which can store temporal features

from sign-language.

2d convnets are used for static getsures, 3d convets can be used to learn spatio-temporal features. RNN's can also be used, but I don't know much about that.

from sign-language.

i see, may i know after i trained a keras model(hdf5), is that possible to deploy it to android phone for recognition? i need to convert keras to tf model?

from sign-language.

Yeah you may need to convert to tf model and use the Tensorflow Java API

from sign-language.

thanks~ i got a problem here, since i need to recognize dynamic gesture, then i might want to store n frames for prediction, i will use keras library (keras.preprocessing.image import img_to_array, load_img) when predicting the gesture, but tf api doesn't have this function. so what should i do for this part?

from sign-language.

whether you use RNNs or LSTMs, how is the dataset going to look like?

from sign-language.

im using lstm, dataset is image frames from video clips

from sign-language.