why gluoncv RecordFileDetection
needs idx files?
isn’t the purpose of .rec
to conveniently pack together data, labels and index?..
Cheers
From the docs:
*.idx file is used to provide random access to the binary file
The .rec format only gives sequential access. Because images can have different sizes you can’t easily access from imgM
to imgN
by jumping a fixed amount of bytes. I guess we could have imagined a format where the *.idx
information is present at the beginning of the file, but as it stands the *.rec
only contains images and label information.