Duplicate polygons in regression mode
We have found a strange behavior: not sure it is limited to regression, but we have seen this only in such context.
We first run the chain with a ground truth containing several times the same polygon, but with a different label/target value. In the configuration file, check_inputs
was set to True
.
Surprisingly Iota2 runs without complaining (the duplicate polygons were sampled only once). If we runs check_database.py
tool, it properly removes the duplicate polygons.
Not sure this is a correct behavior. In classification iota2 usually complains if duplicate polygons are found.