Structure of train_all.pkl

[rsinghal04]

I am using the dataset for clustering and could not understand the structure of cur_glyph['class'] in dataloader.py. I want to extract each glyph and its class labels. Can some guide me?

[Yizhi Wang]

I am using the dataset for clustering and could not understand the structure of cur_glyph['class'] in dataloader.py. I want to extract each glyph and its class labels. Can some guide me?

Hi rsinghal04,

cur_glyph['class'] should be an array of [0, 1, 2, ..., opts.char_categories-1] (opts.char_categories = 52 in our experiments: a-z, A-Z)
cur_glyph is a collection of a single font, for example:
cur_glyph['class'][0] = 0, cur_glyph['rendered'][0] is the image of 'A', cur_glyph['sequence'][0] is the drawing commands of 'A' ;
cur_glyph['class'][26] = 26, cur_glyph['rendered'][26] is the image of 'a', cur_glyph['sequence'][26] is the drawing commands of 'a'.

Naming cur_glyph as cur_font would be better, sorry.
You can print the data and its shape to see its structure.