Open txhai opened 6 years ago
same
It means that when you create the lmdb dataset, you should sort the image according to its text length.
can i think it means train txt like this: [img0.png 1234 img1.png 12345 img2.png xyzwe img3.png 123456788] ,its sort by label length?
Hi @meijieru , Since lmdb stores the data in the dictionary, the sequences based on length is not preserved. Any ideas?
Hi @meijieru , Since lmdb stores the data in the dictionary, the sequences based on length is not preserved. Any ideas?
The images are retrieved based on the order they were written to lmdb dataset.
Could you please explain this?