Hello @JietChoo ,
Deploy model on board requires some basic knowledge of deep learning. Text Recognition has two models. It is too difficult for new.
First stage is detecting the text in image. The result is like the image below.
Then cut image from the image like below.
At last, recognize characters in each cut image. Densenet CTC is a recognize characters model. It need input like pictures above. That is why Densenet CTC only has image demo.
In you log, model infer is successful. The result is the red boxes. However, as a result of input violation of design, its result is wrong.
Considering you are a new. We will add a new text recognition demo in VIM4 KSNN. But it is only a demo, the accuracy will not be very high. It will take about a month.