Most of the time (some tests for it here and here) we expect importers to work successfully with empty datasets. But even if we can export such datasets, in some formats, we cant detect them "back" with detect_dataset method. That happens because detector methods expect annotations not to be empty and throw an error like:
"'/gt.txt' must be a ICDAR-like annotation file"
List of formats:
The solution here may be correcting the detection code so it can accept empty files.
Most of the time (some tests for it here and here) we expect importers to work successfully with empty datasets. But even if we can export such datasets, in some formats, we cant detect them "back" with
detect_datasetmethod. That happens because detector methods expect annotations not to be empty and throw an error like:List of formats:
.txtfile required per each.bmp)The solution here may be correcting the detection code so it can accept empty files.