I am trying to implement and test the approach for video encoding. I would like to have as input to the system sets of image frames from videos and just encode them using only the encoding part. Therefore, I am trying to comment out the decoder part and I am trying to figure out what modifications should I perform to make it work. I am a bit puzzled with the line 30 and 34 in pinyin_main.py:
I am trying to implement and test the approach for video encoding. I would like to have as input to the system sets of image frames from videos and just encode them using only the encoding part. Therefore, I am trying to comment out the decoder part and I am trying to figure out what modifications should I perform to make it work. I am a bit puzzled with the line 30 and 34 in pinyin_main.py:
gen = dd.S2SDataGenerator('data/pinyin.corpus.txt', itokens, otokens, batch_size=32, max_len=120) s2s.model.fit_generator(gen, steps_per_epoch=2000, epochs=5, callbacks=[lr_scheduler, model_saver])Could I replace the gen object with a tensor easily? What exactly gen stands for?