URGENT: What next after running extract_ocr_frcn_feature.py in m4c ?

[Shwet Kamal Mishra]

❓ Questions and Help

I have made some changes in the feature computation of extract_ocr_frcn_feature. This script has created 2 .npy files for each image.
Now, how to use those npy files for training?
Does this need to be used in place of the ocr_en_frcn_features.lmdb folder somehow? If yes, then how to convert the folder with npy files to lmdb?

I am stuck at this step and will highly appreciate any help.
@ronghanghu

[Ronghang Hu]

Hi, you need to use https://github.com/facebookresearch/mmf/blob/master/tools/scripts/features/lmdb_conversion.py to convert your "*.npy" feature files to lmdb. You can follow the procedure in #663 (comment)

[Shwet Kamal Mishra]

Thanks, @ronghanghu. This is helpful.