How can I get the predicted onset and prediected offset about actionformer_release HOT 5 CLOSED

Please refer to Issue 13 for details.

from actionformer_release.

Let me know if there is any further questions. Otherwise, I will mark the issue as resolved.

from actionformer_release.

Let me know if there is any further questions. Otherwise, I will mark the issue as resolved.
another question. I apply your model on UCF101(thumos14) but it works badly with the I3d codes from this link. So could you please give me your I3d codes(for extracting features) and the unprocessed datasets（video, not features） ？

from actionformer_release.

Hi, for THUMOS14 dataset, we use the feature provided by CMCS. The CMCS uses pytorch-i3d-feature-extraction to extract the features. I think the link you mentioned is a I3D ResNet model. and you should pay attention to the feature stride part. For the unprocessed video part, you can refer to the official THUMOS14 website.

Let me know if there is any further questions. Otherwise, I will mark the issue as resolved.
another question. I apply your model on UCF101(thumos14) but it works badly with the I3d codes from this link. So could you please give me your I3d codes(for extracting features) and the unprocessed datasets（video, not features） ？

from actionformer_release.

When using a different version of I3D, the feature dimension / stride / size of local window need to be adjusted accordingly in the config file.

The original issue is about the prediction outputs. If you have questions regarding feature extraction, please open a new issue.

from actionformer_release.