Open shaojun opened 2 years ago
1) The first error (Too many open files
) is a duplicate of this: https://github.com/voxel51/fiftyone/issues/1448
It is resolved using the task_size
parameter: https://voxel51.com/docs/fiftyone/integrations/cvat.html#large-annotation-runs
2) For the second error: The image metadata is necessary when uploading samples to CVAT. It seems that there was an issue computing metadata for 93 of your samples. Check to see if those images were downloaded correctly. You can get the filepaths of those samples with the following:
non_metadata_view = dataset.exists("metadata", False)
filepaths = non_metadata_view.values("filepath")
print(filepaths)
If there is an issue with them, you can delete those images and have FiftyOne redownload them by calling foz.load_zoo_dataset
again for your subset of open images. If not, your choices are to either ignore those samples or attempt to manually populate the metadata for them using ImageMetadata.build_for()
Hi, I'm trying to download a small subset with 2000 images of
Open Images
data of classBicycle
, and then upload the images and annotations toCVAT
.System information
fiftyone --version
): FiftyOne v0.14.4, Voxel51, Inc.Commands to reproduce
this is what I've done:
I then tried to reduce the sample to 600, and
annotate(...)
again, but got another error:Describe the problem
the upload to cvat failed.
What areas of FiftyOne does this bug affect?
App
: FiftyOne application issueCore
: Corefiftyone
Python library issueServer
: Fiftyone server issueWillingness to contribute
The FiftyOne Community encourages bug fix contributions. Would you or another member of your organization be willing to contribute a fix for this bug to the FiftyOne codebase?