Mobile Deep Learning with TensorFlow Lite, ML Kit and Flutter

Testing

Now, in the following steps, we will create functions to predict captions based on the model trained in the previous steps, and test the captioning on a sample image:

1. We're finally at the stage where we can use our model to generate the captions of the images. We create a function that takes in the images and uses the model.predict method to come up with one word at each step until #STOP# is encountered in the predictions. It stops there and outputs the generated caption, as follows:

def predict_caption(img):        in_text = "#START#"    for i in range(max_caption_len):        sequence = [word_index_map[w] for w in in_text.split() if w in word_index_map]        sequence = pad_sequences([sequence], maxlen=max_caption_len, padding='post')         pred = model.predict([img, ...