Datasets:
Tasks:
Feature Extraction
Formats:
imagefolder
Languages:
English
Size:
10K - 100K
ArXiv:
DOI:
License:
Cross-image Object Association GT and Leaderboard
#2
by reyemtm - opened
We have been using this data for cross-image object association. It would be great for the community if the hand-curated LiDAR masks could be converted into ground-truth labels, allowing the dataset to be used for this task as well (we created our own version for this purpose).
On top of that, a label-matching benchmark and leaderboard (speed, GPU/CPU and performance) using the ground truth would be awesome. While cross-frame identity association is well established in the video-ML / multi-object-tracking space, there appears to be much less standardized (open source) benchmarking for cross-image instance association for RTK GPS-tagged oblique/drone imagery.