baudm / parseq

Scene Text Recognition with Permuted Autoregressive Sequence Models (ECCV 2022)
https://huggingface.co/spaces/baudm/PARSeq-OCR
Apache License 2.0
561 stars 125 forks source link

LMDB data sets are the same size #65

Closed yiren556 closed 1 year ago

yiren556 commented 1 year ago

I used the code you provided to create my own LMDB dataset, but found that the memory space occupied by the LMDB format generated by datasets with different image numbers is the same. What is the reason image

baudm commented 1 year ago

Sorry but this is outside the scope of the project. Please check the LMDB docs https://lmdb.readthedocs.io/en/release/#memory-usage