Mobile Deep Learning with TensorFlow Lite, ML Kit and Flutter

Preparing the caption dataset

In the following steps, we will process the loaded captions dataset, and convert it to forms suitable for performing training on it:

1. In this step, we split the image descriptions and store them in a dictionary format for easier use in the future code, as shown in the following code block:

descriptions = {}for x in captions:    imgid, cap = x.split('\t')    imgid = imgid.split('.')[0]        if imgid not in descriptions.keys():        descriptions[imgid] = []        descriptions[imgid].append(cap)

In the preceding lines of code, we broke down each line in the file into the parts of image ID and captions for each of those images. We created a dictionary out of it, where the image ID is the dictionary key and each key-value pair contains ...