Open ryuji0398 opened 2 years ago
I checked the "High_activity_009124" and there are indeed two identical bounding boxes. I'm not sure why it happens; probably caused by a human mistake during annotation.
Updating the published dataset is tricky since it will break the reproducibility of the numbers in the published papers. I will keep that annotation mistake in mind, and if there will be a new release of our dataset, we will fix this mistake.
Thank you for providing this information!
Good catch!
Hi. Thanks for providing the fisheye dataset.
I found a mistake in the CEPDOF annotation. There are duplicate bboxes in the High_activity.json file. It is where the image_id is High_activity_009124. Please delete one of them.
Thank you : )