You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Add option for the YOLOv5 dataset export to add the value for path to the dataset.yaml.
Motivation
A problem occured when working with the pip package of YOLOv5.
YOLOv5 will try before training to resolve the relative train / val inside the dataset.yaml.
Without any reference to the absolute path of the data, it will end up resolving it inside the pip package, not in any project. This will end up in an exception and stopping the training process.
What areas of FiftyOne does this feature affect?
App: FiftyOne application
Core: Core fiftyone Python library
Server: FiftyOne server
Details
By adding the absolute path as value to the key path, this problem can be fixed. (I modified the dataset.yaml after the export to verify this). The already existing variable export_dir could be used for this.
As far as I understood it until now, my proposed changes would be quite small:
YOLOv5DatasetExporter.init(...): Recieves another optional, boolean parameter: export_dir_as_path, default False
YOLOv5DatasetExporter.close(...): If flag has been set, export_dir is added as value to key path
Open for any improvements!
Willingness to contribute
Yes. I can contribute this feature independently.
Yes. I would be willing to contribute this feature with guidance from
the FiftyOne community.
No. I cannot contribute this feature at this time.
The text was updated successfully, but these errors were encountered:
Proposal Summary
Add option for the YOLOv5 dataset export to add the value for
path
to thedataset.yaml
.Motivation
A problem occured when working with the pip package of YOLOv5.
YOLOv5 will try before training to resolve the relative
train
/val
inside thedataset.yaml
.Without any reference to the absolute path of the data, it will end up resolving it inside the pip package, not in any project. This will end up in an exception and stopping the training process.
What areas of FiftyOne does this feature affect?
fiftyone
Python libraryDetails
By adding the absolute path as value to the key
path
, this problem can be fixed. (I modified thedataset.yaml
after the export to verify this). The already existing variableexport_dir
could be used for this.As far as I understood it until now, my proposed changes would be quite small:
export_dir_as_path
, defaultFalse
path
Open for any improvements!
Willingness to contribute
the FiftyOne community.
The text was updated successfully, but these errors were encountered: