search

gurkirt / 3D-RetinaNet

3D-RetinaNet a baseline models on ROAD dataset
Other
74 stars 20 forks source link

Question about testing #7

Closed sapphireh closed 2 years ago

sapphireh commented 3 years ago

I have seen the 'Testing and Building Tubes' part in Introduction which only used for validation data, and it didn't provide method for testing set. To build the 'class VideoDataset' for model test, we must need annotation files, likes 'road_trainval_v1.0.json' used in function '_make_lists_road', however, we don't have the annotation file for testing set. So, is there any ways to just test the data

gurkirt commented 3 years ago

Thank you for your question. All you need is a list of video names and frame list. In line https://github.com/gurkirt/3D-RetinaNet/blob/master/data/datasets.py#L533, return the clip and the rest of the fields can be random numbers but abiding the shapes.

gurkirt commented 3 years ago

Does that help?

GELIELEO commented 3 years ago

I have the same question. Does the test set not need annotations?

gurkirt commented 3 years ago

No, test it private you can only evaluate against it on eval.ai/road challenge page. Alll you need is videos names and list of frames from testset to run evaluation and save the results. Then you can upload those results to eval.ai.