Closed LinJM closed 5 years ago
Hi @LinJM ,
Yes, there is a discrepancy if you directly compare character integer labels with the alphabet indices. It is due to the LabelEncoder from sklearn.preprocessing. The letters are always transformed into the integer labels by means for LabelEncoder. You can find an example case at https://github.com/emreaksan/deepwriting/blob/a1b011563c58f4b94d92b1110fbc98ffaf35c3e5/data_scripts/json_to_numpy.py#L45 or https://github.com/emreaksan/deepwriting/blob/a1b011563c58f4b94d92b1110fbc98ffaf35c3e5/dataset_hw.py#L271
I think the LabelEncoder sorts the alphabet first and then assigns the indices.
Thank you for your response
@emreaksan hello, I found the release dataset char_label's id isn't correspond to the alphabet's id. especially, the first id. Others's id - 33 = alphabet's id.