Hello there 
,
I’m developing a react native app for text recognition. I’m using Tensorflow JS and this Rosetta tflite model.
The prediction array printed in my code, using method array(), is similar to this one:
[[[-25.384523391723633, -28.51205062866211, -24.662263870239258, -26.68767547607422, -24.16399574279785, -22.323328018188477, -23.47882843017578, -22.976106643676758, -26.73689079284668, -24.63434410095215, -24.818147659301758, -21.522701263427734, -23.43208122253418, -23.797555923461914, -26.84471321105957, -23.7003231048584, -23.263507843017578, -21.910234451293945, -21.727222442626953, -21.817363739013672, -26.309913635253906, -23.365047454833984, -26.087039947509766, -24.847457885742188, -25.71664810180664, -24.464893341064453, -27.24349594116211, -27.59737777709961, -26.38694953918457, -21.934438705444336, -22.45738410949707, -23.208606719970703, -26.741500854492188, -25.01622772216797, -26.48637580871582, -23.740406036376953, -28.078567504882812], [-17.181079864501953, -31.371185302734375, -30.681379318237305, -33.88299560546875, -29.09239387512207, -26.231252670288086, -26.744478225708008, -25.66680145263672, -29.72777557373047, -29.213228225708008, -27.65131950378418, -24.920312881469727, -30.8102970123291, -29.32961082458496, -33.37773895263672, ...
I’ve understood that Rosetta tflite does not have a CTC Decoder so I would implement it.
I’m not able to do that! (paste from chatGPT does not help so much)
Is there an example or a javascript pratical guide that could clarify my doubts?
Thank you in advance 