Hello, I have downloaded the latest annotations for SCOD.
But I find fho_scod_test_unannotated.json is about 480MB.
this file contains about 2000000 frames.
Is this file consistent with to test set in the paper?
Yes. To prevent information leakage, additional annotations have been included in the test set. Only the correct annotations from the test set used for the paper will be used to evaluate the final challenge submission.
Hello, we have submitted the results for evaluation, but there are some errors.
Traceback (most recent call last):
File "/code/scripts/workers/submission_worker.py", line 491, in run_submission
submission_metadata=submission_serializer.data,
File "/tmp/tmp731rinei/compute/challenge_data/challenge_1632/main.py", line 91, in evaluate
pred_coco = get_coco_detections(predictions, framesubpath2id)
File "/tmp/tmp731rinei/compute/challenge_data/challenge_1632/main.py", line 29, in get_coco_detections
x = box["x"]
TypeError: string indices must be integers
Also, we have a question: can each image of the prediction result only contain one box prediction?
Could you give me some help? @ego4d-info
Hi, every PNR frame could have multiple box predictions. The bbox under "pnr_frame should be a list of bounding box parameters (x, y, width, height, score).
We also update the challenge page, please ensure your submission follows the format there: