I wanted to pass an original text to generate an image according to the input text. How can i embed it as a char-CNN-RNN embedding?

[PragyanSubedi]

I trained the model and things look great for both the training and test results. However, I wanted to pass my own sentences onto the model to see what it generates. How can I embed my original text (not belonging to either the training or test set) so I can pass it to the model for generation?

I've spent hours trying to find a decent way to create the char-CNN-RNN text embedding for my sentence but have had no luck. Could you please help me out with this @crisbodnar?

Also, I thought about an alternative approach to this since I found no decent solution anywhere:

For example, if i wrote an input sentence that matched partly with the text from the training or test set, I could use the corresponding matched text's pre-trained embedding to generate the picture. Any suggestions for quickly matching my input text with the training/test text to get the corresponding word-embedding?

[crisbodnar]

@PragyanSubedi , see this related issue https://github.com/crisbodnar/text-to-image/issues/8